Duarte da Silva: Ghana bought Africa’s shadow gold. SA wants to jail the man who digs it.

More than 400 tonnes of African gold leave the continent undeclared each year, most of it bound for Dubai. Ghana tried buying it back. It worked, until the central bank lost $1.7bn and sold half its gold reserve. South Africa prefers the fence: a draft law threatening illegal miners with 30 years in jail. Dr Duarte da Silva says it’s aimed at the wrong man. The trafficked digger is easily replaced. The syndicates, dealers and exporters who own the gold are left alone. Meanwhile Harmony spent R748m on security in a single year. So why is the poorest man in the chain the only target?.By Duarte da Silva*.More than four hundred tonnes of African gold leave the continent each year without papers. Ghana tried to buy it back: a state board paying up to ninety-eight per cent of the world price pulled a hundred tonnes into the light in twelve months, and the International Monetary Fund then costed the programme at 22 billion cedis, about 1.7 billion dollars, a sixth of the metal's value. By late September 2026 its exports had become, in the central bank governor's words, "comparably less regular", the board had been ordered to refine before it exported, and on 13 October its chief executive told Parliament that the loss belonged to the central bank, not the board. Ghana's is not the only state gold channel on the continent, and the others have served treasuries, armies and insiders before they served miners. In South Africa the shadow ounce is dug by poor, often trafficked men and owned by armed syndicates, dealers and exporters, while the working mines pay for policing the state does not provide. This paper follows the ounce to its owners and prices the channel that could take it from them. .The shadow trade in African gold has been counted: not less than 435 tonnes left the continent undeclared in 2022, the latest year counted, 405 of them to the United Arab Emirates, worth about 58 billion dollars at the close of 16 October 2026. What was missing was the cost of reversing it, and Ghana has now supplied the first audited answer. Declared small-scale exports rose from 63.6 tonnes to 103 in a year, the mirror image of 2021, when a three per cent tax cut them by ninety-one per cent; the central bank sold half its gold reserve along the way and has not named the off-takers who bought the metal at the far end. South Africa's problem is harder: an illegal trade inside working mines, run by syndicates with inside help, which cost Harmony alone R748 million in security in the year to June 2025, the latest year it has reported, with the figure for the year to June 2026 due on 23 October 2026. Its answer so far is a draft thirty-year sentence for the man underground. .THE FLOW I. What is known, and what was not The shadow ounce has been mapped. Independent auditors, the London bullion market's own country-of-origin returns and the importing states' customs data agree, within a range, on its shape. Artisanal and small-scale miners across the continent produce something like 700 tonnes of gold a year on the Metals Focus count that the London Bullion Market Association uses, a fifth of world mine supply, and most of it leaves without a file. In 2022, the latest year SWISSAID has counted, not less than 435 tonnes were smuggled, 405 of them to the Emirates; over the decade to 2022 the Emirates imported some 2,569 tonnes of undeclared African gold, worth about 115 billion dollars; and in 2024 its imports from Africa rose eighteen per cent to 748 tonnes. Of the artisanal gold that reached the sixty-six Good Delivery refiners in 2023, the latest year the London association has analysed, the total was 44 tonnes, and Ghana's share of it was 0.6 tonnes, against Peru's twenty. At 4,166 dollars an ounce, the close of 16 October 2026, 435 tonnes is worth about 58.3 billion dollars a year, exported without royalty, record or return. What was not known until the last year was the cost of the alternative. Every proposal to reverse the flow, whether a state buyer, an accredited refinery or a certified channel, rests on the assumption that the artisanal miner will sell into the light if the light pays. That assumption has now been tested at national scale in three countries in eighteen months and audited by the Fund in one of them, and the test turned up something the formalisers were not looking for: the people who run these channels, like the people who run the illegal trade they are meant to replace, have interests of their own. Counting the smuggled ounce was never the hard part. The hard part is who pays to bring it back, and who profits either way. EXHIBIT A · THE NATURAL EXPERIMENT II. Ghana's export series, 2018 to 2025 The cleanest evidence that gold follows price through a legal channel comes from Ghana's own export statistics, published by the new gold board on 6 January 2026, and it predates the board by seven years. Declared small-scale exports were 75.7 tonnes in 2018, worth 2.8 billion dollars; 53.4 in 2019; 39.3 in 2020. In 2021 the government imposed a three per cent withholding tax on small-scale gold sales, and declared exports fell to 3.4 tonnes, worth 185 million dollars, a fall of ninety-one per cent in one year from a tax of three points. The gold did not stop being mined; it stopped being declared, and Togo, which mines almost nothing, exported 52 tonnes the following year. In 2022 the tax was cut to 1.5 per cent and declared exports recovered to 22 tonnes; in 2023, 37.4; in 2024, 63.6. In 2025 the tax was removed, the state board began paying up to ninety-eight per cent of the world price, and declared small-scale exports reached 103 tonnes, worth 10.8 billion dollars and, for the first time in the country's history, more than the 96.6 tonnes its large mines exported. No study of artisanal gold has produced a cleaner measure of how the flow responds to price. The Ghanaian miner at the pit edge is not a criminal by disposition. He sells to the best bid. .A three-point tax moved thirty-six tonnes into the dark in a year. A price near spot moved forty back into the light. .EXHIBIT B · THE CHANNEL, BUILT III. What Ghana did, in dates and tonnes The Ghana Gold Board Act, Act 1140, was assented to on 2 April 2025. From 22 April the board took licence applications; by 30 April every foreign buyer had been ordered out of the local trade; from 1 May the board was the sole buyer, assayer and exporter of artisanal gold, dealing outside it was an offence, and licensed buyers were required to pay in cedis at a price set off the Bank of Ghana's reference rate. By 31 May 2026 the board had 1,184 licensed operators: two aggregators, sixty-seven self-financing aggregators, 736 second-tier and 379 first-tier buyers. The deputy finance minister told Parliament on 25 June 2026 what the channel had carried: 135.843 tonnes purchased between January 2025 and May 2026 (135.221 of it small-scale and 0.622 large-scale) for 16.1 billion dollars, 9.8 billion of it in 2025; 104 tonnes of small-scale gold exported in 2025 for more than ten billion dollars; the cedi up forty-one per cent against the dollar in 2025; gross reserves from 8.98 billion dollars in December 2024 to a claimed 13.8 billion a year later, a figure the Fund puts at 11.9 billion. A board-commissioned assessment in January 2026 estimated that the 2024-to-2025 gain of 39.4 tonnes had been "pulled from smuggling", 3.8 billion dollars of new foreign exchange at a benefit-to-cost ratio of eighteen to one. In the first half of 2026 the board bought 54 tonnes, about two and a half a week; by 22 August its chief executive was reporting 170 tonnes and seventeen billion dollars bought through formal channels since January 2025, and a 150-member enforcement task force. Central banks in Tanzania and Burundi visited to copy the model, and the finance ministers of half the continent cited it as a template before its first audited accounts existed. In fourteen months Ghana had built the largest formal channel for artisanal gold on the continent. Two features of the design explain what followed. The board paid in cedis, at a price off the central bank's reference rate, which meant every purchase was a foreign-exchange transaction the central bank had to fund; and it bought everything offered, at every licensed counter, from the first of May, with no cap on volume and no audited accounts due until June 2027. A channel that pays near spot, in local currency, without a volume limit, on a central bank's credit, is a channel whose cost rises as fast as its success. Seventeen billion dollars through a channel that did not exist in April 2025. The cost rose with the success. EXHIBIT C · THE BILL IV. What the Fund found, and what was not disclosed The channel was paid for by the central bank, and in August 2026 the International Monetary Fund published what that had cost. Its board completed the sixth and final review of Ghana's programme on 27 July, and the staff report, reported from 6 August, put the Bank of Ghana's losses under the domestic gold purchase programme in 2025 at 22 billion cedis (about 1.7 billion dollars at the year's average rate, 1.5 per cent of the country's output) and the bank's negative equity at 6.7 per cent of output. The mechanics were specific: the bank bought doré through the board at near-world prices in cedis, sold it to off-takers, and lost about seventeen per cent of the value of the doré between the two; its operating cost on the programme was 14.5 per cent of value in 2025, against a target the Fund set at about five. The bank's own 2025 statements record a foreign-exchange conversion loss of 9.17 billion cedis and 337 million in assay and service fees paid to the board. The Fund's objection was to the programme's "quasi-fiscal characteristics", which it said "could compromise the autonomy of the central bank", and its remedy was the transfer of the programme to the board. The board answers that the loss is the central bank's, not its own: it reports audited 2025 results with an operating surplus of 907 million cedis, on fees of under one per cent of the value it handled. The bank stopped funding the board's purchases in March 2026; a memorandum transferring the programme was signed in July, effective from the first of that month; and from August the board has financed its purchases from commercial banks and off-takers, with audited accounts due by June 2027. The consequences arrived at the pit edge within weeks. Reuters reported on 24 August 2026 that licensed buyers had gone unpaid for up to three weeks and some had halted purchases; the board had raised about 839 million dollars in advances between March and May; a foreign-exchange auction on 3 August raised about 75 million and was paused, with fewer than five banks participating. That is the failure mode every formalisation scheme carries and none had yet demonstrated: let settlement slow, and the pit edge reverts to the man with cash. On 20 August the opposition moved for a parliamentary inquiry seeking "off-taker identities and discount structures"; the Speaker said on 25 August he would admit it. Under a new financing model from 3 August the board reported 1.3 billion dollars of foreign exchange generated in August. Then, on 23 September, the Bank of Ghana's governor told the monetary policy committee that "the pause in gold exports by GoldBod since August" called for "a careful look at our buffers", and that rebuilding reserves would be "a key priority". At the press briefing the next day he said exports "did not pause, but rather became comparably less regular", and that the board had shipped "quite a significant amount" the week before; the board then reported 1.871 billion dollars of foreign exchange generated in September, against a 1.4 billion target, and from 1 September it had barred the export of unrefined doré, requiring refining at an approved Ghanaian refinery first. The central bank's own series shows what the irregular months cost: gross reserves peaked at 14.16 billion dollars in March 2026, fell to 11.07 billion in August, 4.2 months of imports, and had recovered to 12.05 billion by 22 September. Its trade figures for the first half of 2026 show gold exports of 12.5 billion dollars, up forty-nine per cent, on volume down half a per cent at 2.8 million ounces, large mines and small together: the price did the work, and the tonnage had stopped growing. On 13 October the board's chief executive told Parliament's Public Accounts Committee that 80 to 90 per cent of what it exports is now refined in Ghana and that it would exceed its 2025 foreign-exchange target of 10.8 billion dollars; he gave the committee no tonnage for the year. The scheme did not break at the pit edge, where the price held and miners kept selling, or at the refinery gate, where the doré was accepted. It broke in the treasury. Each month the board bought at ninety-eight per cent of the world price in cedis and sold into a market that paid less, and the gap landed on the central bank's balance sheet as a subsidy no budget had voted. The price that wins the flow has to be paid by a balance sheet built to carry it, and a central bank's reserve is not that balance sheet. And the reserve, the point of the exercise in the government's own telling, went into reverse. The bank's gold holdings rose from 30.53 tonnes in December 2024 to 38.04 in October 2025. In the last quarter of 2025 the bank sold about 18.5 tonnes at around 3,500 dollars an ounce, some 2.22 billion dollars, while spot traded between 4,054 and 4,316; holdings ended the year at 18.6 tonnes. The governor called it "rebalancing", on 29 January 2026, because gold had exceeded forty per cent of reserves; the government's 2026-to-2028 plan then proposed repurchasing 18.5 tonnes at about 5,500 dollars. The channel had built a reserve of gold and the bank had sold half of it at a discount of a fifth to fund the channel. Meanwhile the state's fiscal take from the sector it had formalised was, in 2025, less than 500,000 cedis in tax on 10.8 billion dollars of exports, against 19 billion cedis from the whole mining sector, because small-scale mining paid no royalty until a two per cent levy was introduced in March 2026. Between July 2025 and March 2026 the state ran 122 military operations and made 237 arrests against illegal mining; the board's own profiling of the sector, begun in June 2026, reported that the artisanal sector was "far bigger" than the licensed small-scale one, with 700,000 to 800,000 people mining illegally, seventy to eighty per cent unlicensed, in 120 of the country's 261 districts, and about 50,000 of them Chinese. What the accounts do not show matters as much as what they do. The central bank sold the doré to off-takers at a loss of about seventeen per cent of its value, and the off-takers have not been named; the opposition's motion of 20 August 2026 asked for just that, “off-taker identities and discount structures”. On 13 October 2026 the board's chief executive, Sammy Gyamfi, told the Public Accounts Committee that the 22 billion cedis was not in the board's statements and sat on the central bank's books, and reported an overall surplus of about 5.4 billion cedis for the board; the central bank, invited for the same day, was at the Fund's annual meetings in Washington and is to appear later. The committee's chair asked whether the figure was a loss or an accounting treatment. Ten weeks after the Fund's report, the two institutions that ran the channel had still not sat in the same room to say whose loss it was. Hundreds of excavators seized in operations against illegal mining later disappeared without explanation, on reports cited by the Africa Center for Strategic Studies. And on 20 July 2026 a Ghanaian court sentenced Bernard Antwi Boasiako, the New Patriotic Party's chairman for the Ashanti region, to twenty years in prison for granting mineral rights without ministerial approval and knowingly enabling illegal mining. He has appealed and applied for bail pending appeal, which the Attorney-General opposed on 11 August. The channel was built by a political class some of whose members have been in the pits, and the one party to the programme that made money on every ounce, the buyer who took the discount, is still anonymous. The flow followed the price. The central bank paid for it, lost a sixth of the value and sold half its gold. The buyers at the far end have not been named. EXHIBIT D · THE COMPARATORS V. Ethiopia's premium, Tanzania's fifth, and whom the channels served Ghana is not the only test, and the others run the same way. Ethiopia's central bank pays artisanal miners five to fifteen per cent above the international price; gold exports rose from four tonnes in the 2023/24 fiscal year to about thirty-nine in 2024/25, the bank bought 29.4 tonnes in ten months to May 2025, and reserves rose three billion dollars to 4.4 billion. A tenfold jump in one year, from paying over spot: the sharpest price-elasticity datum on the continent after Ghana's. The Fund's fifth review required a plan to phase out the premium by the end of September 2026, private banks admitted and a plan for the central bank's exit from the trade by December; in the fiscal year to July 2026 gold exports earned 5.65 billion dollars on output of 44 tonnes. The same experiment, the same result, the same referee. Tanzania chose a smaller lever and kept it. From 1 October 2024 every gold dealer and refiner must offer twenty per cent of output to the Bank of Tanzania, which bought 418 kilograms in the fiscal year to June 2024 and 21,710 kilograms, seventy-six per cent of the gold sold through the country's five domestic refineries, in the year to June 2026; 28 tonnes in eighteen months to July, on national output of 67.8 tonnes, with more than 4,000 miners and traders opening bank accounts to be paid. Zimbabwe's Fidelity channel took about 31 tonnes in the eight months to August 2026, more than seventy per cent of it from small-scale miners, and targets 50 for the year. Uganda, which mines little, exported 5.8 billion dollars of gold in the year to November 2025, forty-six per cent of all its exports, through nine licensed refineries, and declared 62 tonnes in 2025 against a production "far short" of that on the United Nations' expert count, with the Congolese province of Ituri, next door, declaring half a tonne. On 25 June 2026 the United States Treasury designated Gasabo Gold Refinery in Rwanda, its chairman and general manager, and three mining companies for moving Congolese gold for an armed group, the first refinery on the continent to be sanctioned by name. The Democratic Republic of Congo's state buyer, Primera Gold, exported five tonnes in 2023, was taken fully into state hands and renamed DRC Gold Trading in December 2024, lost its Bukavu branch to the rebel advance in March 2025, and targets fifteen to eighteen tonnes in 2026. Sudan's state company exported 53 tonnes in nine months of 2025 and banned private exports in September, then reversed the ban in November; of the year's 70 tonnes, eighty-five per cent artisanal, only 14.7 tonnes, on the central bank's count, passed through official channels. Mali created a state body for artisanal gold in July 2026, seven months after a 430-million-dollar settlement with its largest miner reopened Loulo-Gounkoto. Nigeria delivered London-standard bars bought in naira from artisanal miners to its central bank in March 2026, whose gold holdings reached 3.5 billion dollars. The comparators also show whom the channels served first. Ethiopia's central bank paid above the world price because it needed dollars, and 12.2 tonnes of what it bought came from Tigray, a region emerging from war. In Zimbabwe, Al Jazeera's “Gold Mafia” investigation of 2023 filmed Uebert Angel, the country's ambassador-at-large, offering to move more than a billion dollars of dirty money under diplomatic cover, and Henrietta Rushwaya, head of a national miners' body and the president's niece, proposing the central bank's Fidelity refinery as a place to deposit cash; the government announced an investigation. Primera Gold, the Congolese state's joint venture with a company registered in the Emirates, held exclusive rights to buy artisanal gold in South Kivu, and the Global Initiative against Transnational Organized Crime found gold sourced from areas held by armed groups and concluded that “it is unclear where the revenues generated by the joint venture went”. In Sudan both sides of the civil war finance themselves with gold, the army through the state mineral company and the Rapid Support Forces through a company called Al-Junaid, with most of the metal going to the Emirates. Mali's military government seized about three tonnes of Barrick's gold in January 2025 and in July flew about a tonne out of Loulo-Gounkoto by helicopter, before the settlement. Formalisation was the stated aim every time. Foreign exchange, war finance, regime revenue and private gain are what the record shows. Set them on one axis and the pattern is the one Ghana drew. Where the channel pays near or above spot, the flow arrives: Ethiopia's tenfold, Ghana's forty tonnes, Tanzania's seventy-six per cent. When the payer is the central bank, the Fund ends it, in Ghana in March 2026 and in Ethiopia by December. A refinery that asks no questions, as in Uganda, Rwanda and the Emirates, draws the flow too, and now draws sanctions; a state that fences instead of buying, as with Congo's collapsed buyer and Sudan's export ban, is simply gone around. The price paid at the pit edge predicts the tonnage. Who pays it predicts whether the channel survives. Who runs it, and whether its accounts are published, predicts whom it serves. .The channel test across the continent, 2025–26. Every figure is sourced and dated in the note; the reading of whom each channel served is the author's. The channel that pays near spot gets the flow, and the central bank that pays for it gets the Fund. Where the accounts are closed, someone else is paid first. .EXHIBIT E · THE CURRENCY, NOT THE GOLD VI. What Ghana was actually buying The Ghanaian numbers bear one more reading, which explains why the state paid what it paid. The board was not, in the government's own telling, buying gold. It was buying foreign exchange. Every ounce purchased in cedis and sold abroad for dollars was a dollar of reserves the country did not have to borrow, and the cedi's forty-one per cent appreciation in 2025, from one of the worst-performing currencies on the continent to the best, is the outcome the programme was built for. On that measure it succeeded. The trouble is that a central bank that pays near-spot in local currency for a commodity and sells it for dollars is running a foreign-exchange intervention through a gold counter, and the Fund objected to a monetary authority conducting fiscal policy with its balance sheet, which is what "quasi-fiscal operations" means, rather than to the gold. The appreciation the programme bought is also what made the programme's losses in cedi terms so large: a bank that buys gold in a strengthening currency and books its losses in that currency loses on the exchange rate it created. The 9.17 billion cedis of conversion loss is, in part, the price of the cedi's own success. That success has since partly reversed: from 10.45 to the dollar at the end of 2025 the cedi had weakened to about 11.7 by early October 2026, nine and a half per cent, as the board's exports became irregular and the central bank's reserves fell three billion dollars from their March peak. That matters for the design, because it separates the two things Ghana bundled. The formalisation of the ounce (assay, custody, a licensed buyer, a documented export) is a trade-regulation objective that a private channel can deliver at five points. The accumulation of reserves is a monetary objective that a central bank can pursue by buying a fifth of refinery output, as Tanzania does, from a channel someone else runs. Ghana asked one institution to do both with one balance sheet, and the institution that polices balance sheets said no. A country that wants the tonnage without the veto has to unbundle them. Ghana was buying dollars through a gold counter, and the cedi rose forty-one per cent. EXHIBIT F · THE PRICE OF ANONYMITY, QUANTIFIED VII. Seventeen per cent, five per cent, and the ten points between The Fund's audit of Ghana gives the continent something it never had: a measured price for outbidding the smuggler, seventeen per cent of the doré's value lost between purchase and sale, at an operating cost of 14.5 per cent against the Fund's target of five. Ethiopia's premium of five to fifteen per cent over spot is the same number from the other side. That is the sovereign ledger. On the market side, the London association's own analysis says that of roughly 700 tonnes of artisanal gold a year, 21 tonnes, three per cent, reached its refiners in the 2020/21 base year, and 44 in 2023. Ghana's 103 declared tonnes of 2025 went almost entirely to refiners outside the London list, since only 0.6 tonnes of Ghanaian artisanal gold reached them. The channel exists. The papers do not. A hundred tonnes of formalised, state-bought, cedi-paid Ghanaian gold is being refined somewhere that does not report to London, and will meet the tenth version of the responsible-sourcing guidance in 2027 without a file. The margin can now be stated as a number. An accredited private channel, in which a licensed aggregator buys against assay, pays in days and feeds a Good Delivery refinery that generates the four documents of provenance as it operates, runs at about five per cent, the Fund's own target. The ten points between that and the sovereign's fifteen to seventeen are the formalisation margin of the continent: enough to pay the miner more than the smuggler, fund the aggregator's assay and custody, carry the refinery's compliance and leave a return, without a central bank's balance sheet and therefore without the Fund's veto. The state's role in that channel is to license, accredit and enforce. Ghana's error was the payer, not the price. Tanzania's twenty per cent rule, which sends a fifth of refinery output to the central bank and leaves the other four-fifths to a private trade the state has licensed, is closer to the design that survives, and it has not drawn the Fund's objection. A sovereign channel at ninety-eight per cent loses a sixth of the metal. An accredited private channel runs at five. The ten points between them are the continent's formalisation margin. EXHIBIT G · THE SHADOW OUNCE AT HOME VIII. South Africa's shadow ounce carries a gun South Africa's version of the problem is harder than Ghana's, and the comfortable account of it is wrong. That account has poor men, many of them retrenched miners, scratching at shafts the industry abandoned. Illegal mining is not confined to abandoned ground. Harmony recorded 2,320 illegal-mining incidents across 31 sites in the year to June 2025, the latest year it has reported; its results for the year to June 2026, published on 27 August 2026, carried no illegal-mining figure, and its annual and sustainability reports for that year are due on 23 October 2026. Sibanye-Stillwater recorded 614 illegal-mining incidents at its South African gold operations in 2025, against 540 in 2024 and 515 in 2017, the previous peak, and 1,438 arrests, on the fact sheet it published with its 2025 reports in April 2026; by 31 August 2026 it counted 6,925 illegal miners arrested on its ground in eight years, 4,164 of them in the last three. Pan African Resources arrested about 4,000 illegal miners at its underground operations in Mpumalanga in the year to August 2025, the latest figure it has given, between 150 and 200 a week, and police detained 494 in a single operation as they came out of its Sheba mine. These are working mines with employees underground, and many of the people entering them are armed. In June 2022 an electrician at a Sibanye-Stillwater operation was killed in a confrontation between mine security and about 150 armed illegal miners. The violence does not stay in the tunnels. On 12 August 2026 police on an anti-illegal-mining operation near Barberton, where Pan African mines, came under fire from armed suspects; one officer was wounded and airlifted to hospital. Eighteen days later six men were found shot dead on a mountainside at Ngodwana, in the same province, in what police called rivalry between illegal mining groups. On 1 October two more bodies were found at Mashobote near Schoemanskloof, with rifle and pistol cartridges around them and an AK-47 magazine and live rounds seized nearby the day before; the gun battles between rival groups had by then closed stretches of the N4, the road to Maputo, and the president of Agri Mpumalanga said farmers' trucks could not use it and that the armed groups needed specialist police rather than the local station. Twenty-two suspected illegal miners were arrested at Mashobodo over the weekend of 3 and 4 October, thirteen of them with crushing tools and the chemicals used to refine gold-bearing material; twelve of the thirteen were Mozambican and one from Eswatini. Executives are not insulated from the country's violence either: in September 2021 Pan African's chief operating officer, Bert van den Bergh, was shot and seriously wounded outside his Johannesburg home, in what the company described as a failed robbery. No link to illegal mining has been reported in that case, and none is claimed here. The men underground are the cheapest part of the business. A study for the Institute for Security Studies' ENACT programme in 2019 described four tiers: some 30,000 illegal miners at the bottom; the syndicate bosses who organise, supply and arm them; buyers, among them licensed scrap-metal dealers, who purchase the gold; and holders of gold export permits who move it abroad. It traced at least 34 tonnes to Dubai between 2012 and 2016. Richard Stewart, then Sibanye-Stillwater's regional chief and now its chief executive, described the trade in March 2023: “These are highly organised syndicates and there's human trafficking involved, forcing people to go underground.” In 2017, the last year for which Harmony has published a breakdown by nationality, 3,066 of the illegal miners it recorded were Zimbabwean, 1,143 from Lesotho, 864 from Mozambique and 397 South African; in the year to June 2025 the police operation in its areas made 2,603 arrests, 2,109 of them of people in the country illegally. Harmony told Parliament in 2020 that it had handled 641 dead illegal miners over the years, and that on one day in 2009 it recovered 78 bodies at a single shaft, from murder, hunger, dehydration and methane.Stilfontein showed who holds the keys. When police cut supplies to a closed mine in North West in 2024, armed gangs underground stopped men from surfacing for fear of what the bosses would do. When a court-ordered rescue began on 13 January 2025, 246 men were brought up alive and 78 bodies recovered; the police count of dead reached 87, and the Police Minister told Parliament 93. The man police described as a kingpin, James Neo Tshoaeli, a Lesotho national known as “Tiger”, came up a shaft under police guard and never reached custody. Four police officers were arrested on 31 January 2025 for helping him escape. An ounce dug illegally still has to be processed, bought, papered and exported, and each step needs someone on the inside. Sibanye-Stillwater has charged 4,172 employees and contractors since 2013 with aiding and abetting illegal miners, 877 of them in 2025, the most in any year; it recorded 729 incidents of employee and contractor collusion in 2025, against 695 in 2024, and reports that the method has shifted from handing clock cards to illegal miners to smuggling them food and supplies in contractor trucks and hidden compartments. Its security staff were assaulted 151 times in 2025. In May 2026 two senior police generals, Major-General Feroz Khan of Crime Intelligence and Major-General Ebrahim Kadwa, head of the Hawks in Gauteng, were arrested with a Durban businessman, Tariq Downes, and charged in connection with illegal dealing in precious metals; they have not been tried and are presumed innocent. In October 2025 Johannesburg's metro police and the Hawks raided a registered precious-metals refinery in Booysens that, on the police account, was “illegally processing gold-bearing material for illegal miners”. And the exit shows in other countries' statistics. Partner countries reported importing more than 240 tonnes a year from South Africa between 2015 and 2023, over 660 in 2015 alone, against a mined output of about 100 and declared exports of 102 tonnes in 2023, a year in which Rand Refinery's own count was 220.6. Much of the gap is foreign gold refined in Germiston and re-exported, and some is paper, as when Chinese customs booked 377 tonnes of South African gold in 2015 of which 314 had been shipped from Hong Kong, 62 from London and none from South Africa. Two series do not wash out that way. The Emirates reported importing about 7 tonnes of gold from South Africa in 2016, 21 in 2017, 28 in 2018, 26 in 2019, 43 in 2020, 41 in 2021, 31 in 2022 and 17 in 2023, some 215 tonnes over the eight years; Rand Refinery, which refines 98 per cent of the country's mined gold, declared exports to the Emirates of under 18 tonnes over the same years, 8 of them in 2020 and none at all in 2021 and 2022, the two years in which the Emirates recorded 72. SWISSAID's reading is that the refinery “manifestly only played a minor role in this bilateral trade, which raises the question of who the exporters were”, and an Emirati trader told it in 2022 that most of the metal was artisanal gold, dug by zama zamas or smuggled in from a neighbour. The second series is scrap. Refiners on the London list recorded 51.9 tonnes of “recycled” gold from South Africa in 2018 and 64.8 in 2019, of which 36.1 and 49.4 tonnes went into Switzerland and 7.2 and 5.1 into China. Metals Focus puts the country's entire scrap market at 25 to 30 tonnes in those two years and about 10 in the years either side; Rand Refinery, which says that “very little secondary material is sourced from within South Africa”, puts it at two to five tonnes a year, and its chief executive allowed in 2021 that much of this scrap may contain illegal miners' gold. The line then collapsed: the Swiss refiners' “recycled” intake from South Africa fell to 8.6 tonnes in 2020, 0.2 in 2021 and 1.7 in 2022, and the whole recorded flow to 8.6 tonnes by 2022. In none of those years did any London refiner record a kilogram of artisanal gold from South Africa. Sixty tonnes a year of “scrap” from a market of ten, and two hundred tonnes of Emirati imports the only accredited refinery did not send, are the shadow ounce's exit, measured from the far side. The working mines pay the bill the state has left them. Harmony's security spending, as it reported to Parliament, rose from R120 million in 2012 to R322 million in 2019 and R678 million in the year to June 2024; in the year to June 2025, the latest year disclosed, it was R748 million, and the company has demolished or sealed 48 shafts since 2008. Its figure for the year to June 2026 is due with its annual reports on 23 October 2026. Sibanye-Stillwater publishes no rand figure for its security spending: its 2025 reports, issued in April 2026, count incidents, arrests, charges and collusion but not the bill, and the most recent figure on the public record remains Stewart's, that security costs were “north of” a billion rand a year by 2022, on top of more than a billion rand of production lost that year to copper-cable theft. Its direct losses to copper theft were R39 million in 2025, against R53.4 million in 2024. Pan African's security cost rose from about 25 dollars to about 40 an ounce by August 2025, and in May 2025 it retrenched 244 people after concluding that theft, including collusion between employees and illegal miners, had made parts of Barberton uneconomic. The state's own figure for the loss, 60 to 70 billion rand a year, has never been reconciled with any export statistic; a 2017 PwC estimate put it at seven billion, and estimates of illegal output run from 7 to 51 tonnes a year. What is not in doubt is who has been paying for the policing. In South Africa the shadow ounce is dug by the poor and owned by syndicates, dealers, refiners and exporters, with insiders in the mines and the police. The working mines pay to keep them out. EXHIBIT H · THE HOME CHOICE IX. The fence, aimed at the wrong man South Africa made its choice in the month the Ghanaian channel was repriced, and it chose the fence. On 7 August 2026 the Justice Department gazetted for comment a General (Mining) Laws Amendment Bill, which the Justice Minister presented on 11 August, that would make illegal mining a crime punishable by up to thirty years' imprisonment and a hundred-million-rand fine, against a maximum of 100,000 rand and two years under the mining act today, and 250,000 rand under the Diamonds Act; comments closed on 11 September, and the Bill had not been tabled by 16 October 2026. The draft defines an "artisanal miner" and an "artisanal mining permit" without saying how such a permit is obtained, and it adds two offences to Schedule 1 of the Criminal Procedure Act, the schedule that governs arrest without warrant and bail; the University of Cape Town's mineral-law programme, on 5 October, called it a war without a strategy, aimed at the miner rather than at the suppliers of explosives and mercury and the markets that buy the gold. Behind it stand the numbers above, and two more: 1,826 illegal miners brought to surface and arrested at Stilfontein, and a police operation estimated to have cost 1.77 billion rand. The draft mineral-resources bill of May 2025 does contain, for the first time, an artisanal tier (permits for operations under five million rand and small-scale permits to fifty million), but it has not reached Parliament, while the criminal bill is already out for comment. The severity matches the violence. The aim does not. Most of those arrested in the mines are undocumented migrants, and arrests of that kind often end in deportation, as hundreds did after Stilfontein; the syndicate recruits replacements. A thirty-year sentence for a trafficked digger removes one worker from a business with a queue of them, and does nothing to the dealer who buys his gold, the refiner who processes it, the permit-holder who exports it or the official who looks away. Neither Ghana nor South Africa has run the design that the numbers above say survives: an accredited private channel, licensed by the state and paying near spot against assay, feeding a refinery that reports to London. South Africa is the one country on the continent that already owns every component of that channel: the assay laboratories, the licensing regime under the Precious Metals Act, and Rand Refinery, the London list's referee since 2003, with some 600 tonnes a year of capacity and sixty-one per cent of its 2024 feed already arriving from outside South Africa, most of it from large African mines. It is also the one country that has reached for a thirty-year sentence before a licence. The lesson that the sovereign cannot be the buyer argues for that design, not against formalisation. South Africa also has the laws to reach the owners of the illegal ounce (the Precious Metals Act's offences for dealing in unwrought gold, the racketeering and forfeiture provisions of the Prevention of Organised Crime Act, the trafficking law that covers the recruiters) and has aimed none of them at the top of the chain as hard as it now proposes to aim a thirty-year sentence at the bottom. Ghana bought the ounce and could not afford it. South Africa fences the ounce and aims its heaviest sentence at the poorest man in the chain. THE BEAR, MET X. What the sceptic rightly takes The bear's first blow is the obvious one: Ghana's channel is failing, and its exports became irregular in August. A state buyer that loses a sixth of the value, sells its reserve at a discount and cannot pay its own agents is not a model; it is a warning. Conceded, and the warning is about the payer rather than the price. The tonnage data show the flow followed the channel; the fiscal data show the central bank could not carry it. The second blow is that the private channel is a theory and Ghana's was a practice; nobody has run an accredited aggregator network at a hundred tonnes a year anywhere in Africa. Conceded as the gap. What exists is the components: Rand Refinery's 600 tonnes, the Swiss "better gold" schemes at a few tonnes, Fairmined and Fairtrade at a tonne or two, Tanzania's licensed refineries at 28 tonnes, and the Fund's own target of five per cent as the cost at which a channel is allowed to survive. Then the file. Even a private channel at five points does not make Ghanaian gold Good Delivery gold unless the refinery reports to London, and the 103 tonnes of 2025 went somewhere that does not. Conceded, and it is the reason the programme below ends at a Good Delivery gate rather than at a bank counter: a channel that formalises the ounce without documenting it has moved the metal from the smuggler to the grey market, and from 2027 the grey market's discount begins to be priced. Then the smugglers adapt. Togo's 52 tonnes, Uganda's 62 and Rwanda's refinery show that every channel that pays creates a new border to smuggle across, and Ghana's 170 tonnes may simply be Burkina Faso's and Côte d'Ivoire's gold arriving for the price. Partly conceded: the board's own profiling found the artisanal sector "far bigger" than the licensed one, and some of the flow is regional. But regional arbitrage into a channel that assays, records and pays in a documented currency is formalisation by another name, and the sanction on Gasabo shows what happens to the channels that do not. And South Africa's problem is different. Its illegal mining is underground, armed and syndicated, much of it inside working mines, and a licence tier will not reach a syndicate's man at 1,500 metres. Conceded, and it is why the programme below runs on two tracks: a licensed channel for artisanal gold on the surface and on abandoned ground, and enforcement aimed at the owners of the armed trade underground. Last, the mining houses exaggerate to get the state to police their property. Partly conceded: the 60 to 70 billion rand headline has never been reconciled with export data. The companies' security bills, arrest counts and retrenchments are another matter; they are reported to shareholders and to Parliament. The bear takes the state buyer, the missing file, the leakage across borders and the size of the headline loss. He leaves the price test, the ten-point margin, the unnamed buyers and the syndicate. THE PROGRAMME XI. The channel that survives the Fund, and the law that reaches the owners The programme is a design, and it is drawn from the failures. The payer is not the central bank. A licensed private aggregator buys at the pit edge against assay, at a posted price near spot, in the local currency, and settles in days. Settlement discipline, which Ghana lost in August 2026, is the whole of the miner's trust, and it belongs on a commercial balance sheet that lives or dies by it. On price, the Fund's five per cent is the operating target, which leaves the ten-point margin to pay the aggregator, the assay and the compliance; the state's take is a royalty at Ghana's two per cent, collected at export, not a tax at the pit that the 2021 experiment showed collapses the flow. At the gate, every ounce the channel buys ends at a refinery on the London list, so that the aggregator's assay, custody and origin records become a file the tenth guidance, due in final form in December 2026 and implemented through 2027, will accept: Rand Refinery for southern Africa, and the association's proposed artisanal Good Delivery lists for the rest. The state's role is to license the aggregators, accredit the laboratories, enforce against the unlicensed buyer with the criminal law the Cabinet has just drafted, and keep the reserve purchases to a Tanzanian fifth of refinery output rather than a Ghanaian whole. And South Africa should go first: table the artisanal tier of the May 2025 bill alongside the criminal bill, confine it to surface and abandoned ground, and let the one country on the continent that owns the refinery license the channel its neighbours are paying 1.7 billion dollars a year to improvise. Who acts: the Minister of Mineral and Petroleum Resources licenses the aggregators and tables the tier; the South African Diamond and Precious Metals Regulator accredits the laboratories; Rand Refinery and the London list are the gate. .Read more:.Duarte da Silva: Joburg's mine dumps hold $215bn of gold, two-thirds is on nobody's books.At home the channel needs a second track, because no licence reaches an armed syndicate inside a working mine. Follow the money: charge the financiers, dealers, refiners that process illegal material, export-permit holders and the officials who protect them under the racketeering provisions of the organised-crime law, and seize the proceeds. Close the exit: audit every dealer's licence, refining licence and export permit issued under the Precious Metals Act, and require a refiner's assay and origin record on every export of unwrought gold. Put the state back on the perimeter of working mines, which the companies have been paying to police, and seal the derelict shafts that serve as entry points. Prosecute the recruiters who send men underground under the trafficking law, and screen rescued miners as possible victims before processing them as offenders. Amend the criminal bill so that penalties rise with a person's place in the chain. Who acts: the National Director of Public Prosecutions, with the Financial Intelligence Centre and SARS; the Minister of Mineral and Petroleum Resources, through the regulator; and the Ministers of Police, Finance and Justice. THE VERDICT XII. The seller, the payer and the owner For a decade the argument about African gold has been about the seller: whether the man at the pit edge would sell into the light if the light paid. Ghana answered it both ways, out of the light under a three-point tax in 2021 and back into it at ninety-eight per cent in 2025. The seller was never the problem. The argument that begins now is about the payer, and the data on that are as clean: a sovereign that pays ninety-eight per cent loses a sixth of the value, sells the reserve it was building, loses its central bank's funding within a year and sees its exports falter within eighteen months. The channel works. The sovereign cannot afford to be the channel.The ten points between a state channel and an accredited private one pay for settlement on the promised day, an assay the seller can check, custody and a refinery that reports, which a central bank cannot supply and a licensed trade cannot survive without. Ghana's board had the price and lost the settlement discipline within fifteen months, because it was spending someone else's balance sheet. The third question is the owner, and it is the one the continent's channels have avoided. In Ghana the buyers who took the discount are unnamed. In Zimbabwe, Congo, Sudan and Mali the channels served insiders, armies and juntas first. In South Africa the illegal ounce belongs to syndicates, dealers, refiners and exporters, with insiders in the mines and the police, and the working mines spend hundreds of millions of rand a year to keep them out. My conclusion is this. The design that survives is a licensed private channel at about five points, feeding a refinery that reports to London, with the state licensing and policing rather than buying; a state buyer left unaudited serves itself. South Africa owns every part of that channel and has not built it, and it has the laws to reach the owners of the illegal ounce and has aimed them at the man at the bottom. It should build the channel and re-aim the law before the London guidance puts a discount on every ounce without a file. .THE SHADOW OUNCE FOLLOWS THE CHANNEL THAT PAYS. GHANA PROVED IT BOTH WAYS. THE SOVEREIGN THAT PAID LOST A SIXTH OF THE METAL, AND THE BUYERS WHO TOOK THE DISCOUNT ARE STILL UNNAMED. IN SOUTH AFRICA THE SHADOW OUNCE IS DUG BY THE POOR AND OWNED BY SYNDICATES, DEALERS AND INSIDERS. BUILD THE LICENSED CHANNEL, AND AIM THE LAW AT THE OWNERS. .*Dr Duarte F da Silva • Capital Markets Strategist • Managing Director, Northbound Processing • Germiston, October 2026 .A NOTE ON THE FIGURES Every figure in this paper was checked against its source in the week of 7 September 2026, re-checked on 23 September 2026, checked again on 26 September 2026, when South African and continental evidence was added, checked again on 11 October 2026, when prices moved to the 9 October 2026 close, again on 13 October 2026, when Sibanye-Stillwater's 2025 figures were added, and again on 17 October 2026, when prices moved to the 16 October close and the Ghanaian parliamentary hearing, the Bank of Ghana's reserve and trade series and the Mpumalanga evidence of October were added; the sources follow. Continental flow: ≥435 t smuggled in 2022 (~US$31bn), 405 t to the UAE, ~2,569 t / ~US$115bn over 2012–2022, undeclared ASM 321–474 t/yr (SWISSAID, "On the Trail of African Gold", 30 May 2024); UAE imports from Africa 748 t in 2024, +18% (SWISSAID, 4 November 2025); 44 t of ASM gold at the 66 Good Delivery refiners in 2023 against 1,824 t large-scale; Ghana ASM 0.6 t vs Peru 20 t at GDL refiners (LBMA Country of Origin Analysis Report 2025); ~700 t/yr of ASM gold, 21 t (3%) reaching GDL refiners in 2020/21 (LBMA, "Towards an LBMA Good Delivery List for ASM"); Metals Focus ASM ~751 t; 435 t × 32,150.7 × US$4,166 ≈ US$58.3bn (author's arithmetic). Ghana: Gold Board Act 2025 (Act 1140) assented 2 April 2025; licences from 22 April; foreigners out by 30 April; sole buyer from 1 May 2025 (GoldBod, 14 April 2025); export series 2018–2025 — 75.7 t (US$2.8bn), 53.4, 39.3, 3.4 (US$185m, after the 3% withholding tax), 22 (tax cut to 1.5%), 37.4, 63.6, 103 t (US$10.8bn, tax removed); large-scale 96.6 t (US$9.2bn) in 2025 (GoldBod, 6 January 2026); 135.843 t purchased January 2025–May 2026 (135.221 ASM, 0.622 LSM), US$16.1bn (US$9.8bn in 2025), 104 t ASM exported 2025 (>US$10bn), cedi +41%, gross reserves US$8.98bn → US$13.8bn, 1,184 licensed operators at 31 May 2026 (Deputy Finance Minister Thomas Nyarko Ampem to Parliament, Graphic Online, 25 June 2026; the IMF puts end-2025 reserves at US$11.9bn); 39.4 t "pulled from smuggling", ~US$3.8bn, benefit-cost ~18:1 (GoldBod-commissioned assessment, 4 January 2026; GBC, 12 January 2026); H1 2026 purchases 54 t, ~2.5 t/week (Reuters via CNBC Africa, 14 July 2026; Citi, 15 July 2026); 170 t / US$17bn bought since January 2025, 150-member task force (Sammy Gyamfi, GhanaWeb, 22 August 2026); GoldBod audited 2025 operating surplus GH¢907m and fees ~0.76% (GoldBod, 19 and 24 August 2026); FX generated US$1.315bn in August under the new financing model from 3 August (GoldBod, 31 August 2026); exports "paused since August" (BoG Governor Asiama, MPC opening, 23 September 2026; Reuters via Mining Weekly), "did not pause, but rather became comparably less regular" (Asiama, MPC press briefing, 24 September; GNA, 25 September 2026); doré export ban from 1 September 2026 (GoldBod directive, 24 August; Reuters, 25 August 2026); September FX US$1.871bn, BoG exit from FX intermediation in October (GoldBod, 30 September 2026; GNA, 1 October 2026); Public Accounts Committee hearing of 13 October 2026 — GoldBod CEO Sammy Gyamfi: the GH¢22bn is not in GoldBod's statements and sits with the BoG, overall surplus ~GH¢5.4bn, 80–90% of exports now refined locally, 2025 FX target of US$10.8bn to be exceeded; chair Abena Osei Asare's question whether the figure is a loss or an accounting treatment; BoG absent at the IMF–World Bank meetings and to appear later (Citi Newsroom, GhanaMMA and GoldBod statement, 13 October 2026); BoG gross reserves US$14.16bn (March 2026), US$11.07bn (August, 4.2 months of imports), US$12.05bn (22 September); H1 2026 gold exports US$12.50bn (+49.0%) on 2.80 Moz (−0.5%) at an average US$4,463.80/oz; cedi GH¢10.45/US$ at end-2025 to ~11.7 in early October 2026 (Bank of Ghana data as reported, Rio Times, 6 October 2026); BoG gold 30.53 t (December 2024), 38.04 t (October 2025), 18.6 t (December 2025), 20.8 t (March 2026), 24.4 t (June 2026) (GoldBod, 5 November 2025; TV BRICS/GNA, 26 July 2026); Q4 2025 sale of ~18.5 t at ~US$3,500 (~US$2.22bn) with spot US$4,054–4,316; Governor Asiama's "rebalancing", 29 January 2026; GHANRAP 2026–28 repurchase at ~US$5,500 (Citi, 29 January 2026; GhanaWeb, 28 March 2026); IMF Executive Board completed the sixth ECF review on 27 July 2026 (IMF press release); staff-report figures reported from 6 August — BoG losses GH¢22bn ≈ US$1.7bn (some outlets US$1.9bn) ≈ 1.5% of GDP in 2025, negative equity 6.7% of GDP, trading losses ≈17% of doré value, operating cost 14.5% (2025) vs ~5% target, "quasi-fiscal characteristics … could compromise the autonomy of the central bank" (Ecofin, 6 August 2026); BoG FX conversion loss GH¢9.17bn and GH¢337.4m in fees (Citi, 26 August 2026); BoG stopped funding March 2026, MoU July 2026 effective 1 July, self-financing from August, audited accounts due June 2027 (MyJoyOnline, 1 September 2026); buyers unpaid up to three weeks, ~US$839m in advances March–May, FX auction ~US$75m on 3 August then paused, fewer than five banks (Reuters/Mining Weekly, 24 August 2026); opposition inquiry motion, 20 August 2026, Speaker to admit it, 25 August (Rio Times, 23 August 2026; GBC); ASM tax revenue <GH¢500,000 in 2025 vs GH¢19bn from mining, 2% royalty from March 2026, 122 operations and 237 arrests July 2025–March 2026, Tanzanian and Burundian central-bank visits (Ecofin, 3 September 2026); galamsey 700,000–800,000 miners, 70–80% unlicensed, ~50,000 Chinese, 120 of 261 districts (Africa Center for Strategic Studies, 21 July 2026); miner profiling from June 2026 (MyJoyOnline, 10 August 2026). Ethiopia: NBE premium 5–15% above spot; exports 4 t (2023/24) → ~39 t (2024/25); 29,396 kg bought in ten months to May 2025 (Tigray 12,210 kg); reserves +US$3bn to US$4.4bn; IMF phase-out plan by end-September 2026 and exit plan by end-2026 (The Reporter, 31 January, 14 and 18 July 2026; allAfrica, 27 May 2025); 2025/26 (to 7 July 2026) gold exports US$5.65bn, output 44 t (Xinhua, 8 August 2026). Tanzania: 20% rule from 1 October 2024 (Mining.com, 28 September 2024); BoT 418 kg (FY2024), 21,710 kg = 75.64% of refinery-sold gold (FY2025/26), output 67.8 t, five refineries, 28 t in 18 months, >4,000 new bank accounts (TanzaniaInvest, 13 August 2026; Capmad, 13 July 2026). Zimbabwe: Fidelity 26.05 t January–July 2026, ASM 70.6% (Zimbabwe Situation, 14 August 2026); August 5.12 t, ASM 77.7%, ~31.2 t January–August against a 50 t target (Mining Zimbabwe, September 2026). Uganda: exports US$5.8bn, 46% of all exports, year to November 2025; nine refineries; 62 t declared in 2025 vs production "far short"; Ituri ~0.5 t (Monitor, 9 January 2026; UN Group of Experts via Africanews, 2 July 2026). Rwanda: OFAC designation of Gasabo Gold Refinery, its chairman and general manager and three mining companies, 25 June 2026 (US Treasury). DRC: Primera Gold >5 t / >US$300m in 2023, fully state-owned and renamed DRC Gold Trading SA, 6 December 2024, Bukavu branch seized March 2025; target 15–18 t in 2026 (Global Initiative, 28 January 2026; Bankable, 14 January 2026; CKM). Sudan: 53 t in nine months, private exports banned 15 September 2025 and reversed 5 November 2025, 70 t in 2025 >85% artisanal, 14.7 t through official channels (central bank) (Sudan Tribune, 12 October and 28 December 2025, 27 February 2026). Mali: state artisanal body, July 2026 (Reuters, 2 July 2026; Africanews, 3 July 2026); Barrick settlement of CFA 244bn ≈ US$430m, 24 November 2025, control restored 18 December 2025 (Bloomberg via Africanews, 25 November 2025; Ecofin, 6 February 2026). Nigeria: LBMA-standard bars to the CBN, holdings US$3.5bn (Vanguard, 5 March 2026). South Africa: General (Mining) Laws Amendment Bill gazetted for comment 7 August 2026 (GG 55166, GoN 4088) and presented 11 August, up to 30 years / R100m (from R100,000 under the MPRDA and R250,000 under the Diamonds Act), comments to 11 September 2026; no tabling record found at 16 October 2026 (SABC; allAfrica, 11 August 2026); the draft's artisanal definitions without permit procedure and its two new Schedule 1 offences (UCT MERLiA, B. Dladla, "The War without Strategy", 5 October 2026; Daily Maverick, "Blunt-force justice", 30 September 2026); ~30,000 illegal miners, 1,826 surfaced at Stilfontein (Miningmx, 31 August 2026); draft Mineral Resources Development Bill of 20 May 2025 (artisanal ≤R5m, small-scale R5–50m); Rand Refinery ~450–600 t/yr, LBMA referee since 2003, 61% of FY2024 feed from other countries, mostly African large-scale mines (African Gold Report). LBMA Responsible Gold Guidance v10: first consultation 10 June–10 July 2026, second consultation Q4 2026, final publication December 2026, implementation through 2027 (LBMA). Fairmined and Fairtrade premia and volumes as previously published. Gold: US$4,166 close on 16 October 2026 (Trading Economics; USAGOLD US$4,165.61); USD/ZAR 16.51 (16 October 2026 close). The "ten points" between a sovereign channel at 15–17% and an accredited channel at the Fund's ~5% target is the author's framing of the IMF figures. South Africa, operating mines: Harmony FY2025 (year to June 2025) — 2,320 illegal-mining incidents at 31 sites, R748m security spending (FY2024: R678m; IOL, 30 October 2025), 48 shafts demolished or sealed since 2008, 2,603 Vala Umgodi arrests in its areas, of which 2,109 were illegal immigrants, 1,406 extracted at Margaret shaft (Harmony, illegal-mining disclosures); Harmony security spending R120m (2012), R166m, R164m, R191m, R215m, R245m, R278m, R322m (2019), 641 deceased illegal miners handled, 78 bodies recovered at Eland shaft on one day in 2009, nationalities of illegal miners in 2017 (Zimbabwe 3,066; Lesotho 1,143; Mozambique 864; South Africa 397) (Harmony briefing to the Portfolio Committee on Mineral Resources, 28 October 2020, PMG). Sibanye-Stillwater: 614 incidents, 1,438 arrests and 877 employees and contractors charged at the SA gold operations in 2025 (2024: 540, 1,487 and 737; 2017: 515 incidents); 3,851 incidents, 9,675 arrests and 4,172 charged over 2013–2025; 729 collusion incidents (2024: 695); 151 assaults on security staff (2024: 141); direct copper-theft losses R39m (2024: R53.4m); no rand security-spend figure disclosed (Sibanye-Stillwater, “Combatting illegal mining” fact sheet, 2025 reporting suite, April 2026, pp. 2, 3 and 5; Sustainability Report 2025, p. 85); electrician killed in a confrontation with ~150 armed illegal miners, June 2022; security costs “north of” R1bn a year and >R1bn of production lost to cable theft in 2022, and the human-trafficking quotation (Richard Stewart, Daily Maverick, 28 March 2023); Stewart chief executive from 1 October 2025; 6,925 arrests on its ground in eight years, 4,164 in the last three (Miningmx, 31 August 2026). Pan African: ~4,000 illegal miners detained in a year, 150–200 a week, 494 at Sheba, security cost up from ~US$25 to ~US$40/oz, 244 retrenchments in May 2025 (Miningmx and Business Day, August 2025); Barberton shoot-out, officer wounded, 12 August 2026 (The Citizen; Lowvelder, 14 August 2026); six men found shot dead at Ngodwana, 30 August 2026, suspected rivalry between illegal mining groups (Sowetan; EWN); two bodies at Mashobote near Schoemanskloof, cartridges and an AK-47 magazine, N4 closures, Agri Mpumalanga's Lion du Plessis (EWN, 1 and 2 October 2026); 22 arrests at Mashobodo on 3–4 October 2026, 13 of them on 4 October with pendukas and refining chemicals, 12 Mozambican and one from Eswatini (The Citizen, 6 October 2026); Bert van den Bergh shot outside his home, 19 September 2021, described by the company as a failed robbery (TimesLIVE, 20 September 2021). The chain: ISS/ENACT, Alan Martin, “Solving South Africa's violent and costly Zama Zama problem”, 19 June 2019 (30,000 miners; bosses, buyers incl. licensed scrap dealers, export-permit holders; ≥34 t to Dubai 2012–16; R14bn output, R21bn losses). Stilfontein: rescue from 13 January 2025, 246 survivors and 78 bodies; 87 dead on the police count (AP, January 2025) and 93 on the Police Minister's reply to Parliament (Daily Maverick, 18 March 2025); armed underground gangs prevented resurfacing (contemporaneous reporting); James Neo Tshoaeli (“Tiger”), four officers arrested 31 January 2025 for aiding his escape (SAPS via Briefly, February 2025); deportations after Stilfontein (IOL, 10 April 2025). State costs: R60–70bn a year claimed, never reconciled with export data (Miningmx, 31 August 2026); PwC 2017 estimate ~R7bn; Vala Umgodi ~R1.77bn (author's estimate from published deployment costs; BizNews). Ghana, additional: excavators seized and later missing (Africa Center for Strategic Studies, 21 July 2026); Bernard Antwi Boasiako, NPP Ashanti regional chairman, 20 years, 20 July 2026, appeal and bail application pending (Miningmx; The Africa Report). Continent: Al Jazeera Investigative Unit, “Gold Mafia”, March–April 2023, and Zimbabwe's announced investigation (Al Jazeera, 5 April 2023); Primera Gold DRC SA, state–UAE joint venture, exclusive South Kivu rights from December 2022, >5 t and >US$300m in 2023, state takeover and renaming December 2024 (Global Initiative against Transnational Organized Crime, “Feeding the furnace”; CKM); Sudan — SAF-area production 64.36 t in 2024, RSF via Al-Junaid, exports mainly to the UAE (Chatham House, “Gold and the war in Sudan”, March 2025); Mali — ~3 t seized January 2025, ~1 t (US$117m) removed by helicopter 10 July 2025 (Mining.com; The Northern Miner), settlement CFA 244bn ≈ US$430m, November 2025. South African law: Precious Metals Act 37 of 2005; Prevention of Organised Crime Act 121 of 1998; Prevention and Combating of Trafficking in Persons Act 7 of 2013; draft Mineral Resources Development Bill, 20 May 2025 (artisanal tier). Insiders and exit: Major-Generals Feroz Khan and Ebrahim Kadwa and businessman Tariq Downes arrested and charged over alleged illegal precious-metals dealings, May 2026 (EWN, 10 May 2026; IOL, 11 May 2026), SAFTU quotation (IOL, 11 May 2026); case postponed to 27 November 2026 (Kempton Park Magistrate's Court, 2 September 2026); Booysens registered refinery raided by JMPD and the Hawks, 29 October 2025, R20,000 alleged bribe, Fortein quotation (The Citizen); partner-country imports >240 t/yr from South Africa 2015–23 (>660 t in 2015), declared exports 102 t in 2023 vs Rand Refinery 220.6 t, Chinese customs 377 t in 2015 (314 t via Hong Kong, 62 t via the UK, none direct); UAE-reported imports from South Africa (UN Comtrade HS 7108, as reported by the UAE and uncorrected for transit) ~7, 21, 28, 26, 43, 41, 31 and 17 t for 2016–2023 (~215 t) against Rand Refinery's declared exports to the UAE of ~0.5, 2, 3, 1, 8, 0, 0 and 3 t (~18 t), both read to the nearest tonne from SWISSAID's chart; “manifestly only played a minor role” and the Emirati trader's 2022 reading; LBMA country-of-origin returns of “recycled” gold from South Africa — total 51.9 t (2018), 64.8 t (2019), 11.2 t (2021), 8.6 t (2022); Swiss refiners 36.1, 49.4, 8.6, 0.2 and 1.7 t (2018–2022); Europe and Africa 8.6 and 10.3 t; China 7.2 and 5.1 t (2018–19); no ASM gold from South Africa recorded in any year; scrap market Metals Focus ~10 t (2016–17, 2021–22) and 25–30 t (2018–19), Rand Refinery 2–5 t/yr and “very little secondary material is sourced from within South Africa”, CEO on scrap containing illegal miners' gold (Engineering News, 2021); Rand Refinery refines 98% of South Africa's industrial gold, 39% of FY2024 feed (90.0 t) South African; illegal and artisanal output estimates 7–51 t/yr and ~25 t (African Gold Report, South Africa country profile, SWISSAID, read 17 October 2026). The four-tier reading of the chain follows ENACT; the reading of each state channel's first beneficiary is the author's. Currency of the South African company figures (checked 16 October 2026): Harmony's latest published illegal-mining and security figures are for the year to June 2025 (its results for the year to June 2026, 27 August 2026, and its Mining Forum Americas presentation of 28 September 2026 carried no illegal-mining figures; its annual and sustainability reports for that year are scheduled for 23 October 2026); Sibanye-Stillwater's latest incident, arrest and charge figures are for calendar 2025 (fact sheet, April 2026) and its latest arrest tally runs to August 2026 (Miningmx, 31 August 2026); its 2025 integrated, sustainability and financial reports were read for this check and none discloses a rand security-spend figure, so the “north of R1bn” figure (2022) remains the most recent on the public record; Pan African's latest figures are those it gave in August 2025. Harmony's 2017 nationality breakdown is the last it has published. The 2022 smuggling count (SWISSAID, May 2024, reconfirmed on its data page of June 2026) and the 2023 LBMA country-of-origin analysis (published September 2025) are the latest available. Northbound Processing • Germiston, October 2026. The Gold Papers — Paper 32. This article does not constitute investment or legal advice. The Ghanaian figures are those of the Gold Board, the Bank of Ghana, the Deputy Finance Minister and the International Monetary Fund as reported and cited, and some are contested in Ghana's own press; the comparator figures are the respective central banks', refiners' and investigators' as cited; the South African loss figure of R60–70bn a year has never been reconciled with export statistics and is labelled as claimed; company security and arrest figures are the companies' own. The shooting of a Pan African executive in 2021 is recorded as the company described it, with no link to illegal mining asserted, and the police officers arrested in May 2026 are presumed innocent. The framing of the ten points as the formalisation margin, and the reading of whom each state channel served, are the author's own. The author runs a business in the metals-recovery economy.

Duarte da Silva: Ghana bought Africa’s shadow gold. SA wants to jail the man who digs it.

Related Stories

Dr Duarte da Silva: The Platinum corollary

Heystek: The tide’s gone out on the JSE – and SA Inc forgot its swimming trunks

Duarte da Silva: SA's richest goldmine is the bin — and almost nobody's mining it

Duarte da Silva: The country we declared explored