Analysis

The Race to the Regulators: Why AI Pre-Deployment Testing Has Arrived

Published

2 months ago

May 17, 2026

For most of the past two years, the dominant assumption in Washington’s corridors was that the Trump administration would keep its hands off frontier AI. The January 2025 revocation of Biden’s executive order on AI risk seemed to cement that posture. So when the U.S. Department of Commerce’s Center for AI Standards and Innovation announced on May 5, 2026 that it had signed formal agreements with Google DeepMind, Microsoft, and Elon Musk’s xAI — granting federal evaluators access to unreleased AI models — the pivot was sharper than most observers had anticipated.

The catalyst was not abstract policy debate. It was a model.

When security researchers at Mozilla pointed Anthropic’s new Mythos system at their code, the experience produced something close to vertigo. Bobby Holley, Firefox’s chief technology officer, said Mythos had elevated AI from a competent software engineer to something resembling a world-class, elite security researcher. That description — and its implications for every unpatched vulnerability in every network connected to the internet — lit a fire under the White House that no deregulatory talking point could easily extinguish. The Washington Post

The new AI pre-deployment testing agreements are Washington’s answer. They are voluntary, technically non-binding, and carefully constructed to avoid the language of mandates. They are also, in their quiet way, a structural reckoning with just how consequential the next generation of AI models may be.

What the CAISI Agreements Actually Do

The Center for AI Standards and Innovation announced agreements with Google DeepMind, Microsoft, and Elon Musk’s xAI that will allow the U.S. government to evaluate artificial intelligence models before they are publicly available. CAISI will conduct pre-deployment evaluations and targeted research. The announcement builds on earlier partnerships struck with OpenAI and Anthropic in 2024, which were the first of their kind. CNBC

The scope is broader than a checkbox exercise. CAISI has completed more than 40 evaluations to date, including assessments involving unreleased AI models. Developers frequently provide models with reduced or removed safeguards to support evaluations focused on national security-related capabilities and risks. The agreements also support testing in classified environments and enable participation from evaluators across government agencies through the TRAINS Taskforce, a group of interagency experts focused on AI-related national security issues. Executive Gov

That last point matters. A model tested with its guardrails intact tells evaluators relatively little about what it’s genuinely capable of doing. By examining systems in their more uninhibited state, CAISI can probe for the kinds of capabilities — automated cyberattack sequencing, biochemical synthesis guidance, manipulation of critical infrastructure — that frontier labs are increasingly warning about in their own internal research.

CAISI’s evaluations focus on demonstrable risks, such as cybersecurity, biosecurity, and chemical weapons. These aren’t theoretical threat categories. They are the precise domains in which advanced reasoning models have begun to demonstrate capabilities that, even in controlled settings, have prompted unusual candour from the labs building them. National Institute of Standards and Technology

Prior to evaluating U.S.-based AI models, CAISI recently examined the Chinese model DeepSeek, concluding it underperformed in several areas including accuracy, security and cost efficiency. That context is not incidental. Part of what’s driving Washington’s urgency is the competitive dimension — the fear that adversaries may be racing toward capabilities that American agencies don’t fully understand, even in their own country’s frontier models. Nextgov.com

CAISI Director Chris Fall has framed the institutional mission with deliberate precision. “Independent, rigorous measurement science is essential to understanding frontier AI and its national security implications,” Fall said. “These expanded industry collaborations help us scale our work in the public interest at a critical moment.” Federal News Network

What Does CAISI’s AI Pre-Deployment Testing Actually Involve?

CAISI conducts pre-release evaluations of frontier AI models by accessing versions with reduced or removed safety filters, testing in classified environments, and deploying an interagency task force — the TRAINS Taskforce — across government agencies. Evaluations focus on cybersecurity, biosecurity, and chemical weapons risks. The center has completed over 40 such assessments to date.

That question has real commercial stakes attached to it. NIST said the partnerships would help the agency and the tech companies exchange information, spur voluntary product improvements, and ensure the government had a clear understanding of what AI models were capable of doing. For the companies involved, this framing is tolerable — even attractive. A pre-release government endorsement, implicit or explicit, is worth something in enterprise procurement conversations. It’s harder to challenge a model that CAISI has already looked at. Cybersecurity Dive

Yet the capacity problem is glaring. CSET Senior Research Analyst Jessica Ji noted that government agencies simply don’t have the same amount of resources as big tech companies — either the manpower, technical staff, or access to compute — to run rigorous evaluations of these models. CAISI is a relatively lean organisation operating against labs that employ thousands of the world’s most skilled AI researchers. The asymmetry between evaluator and evaluated has no obvious near-term solution. CSET

The FDA Analogy — and Why It’s Both Tempting and Dangerous

The policy frame that has seized Washington’s imagination is, perhaps inevitably, the Food and Drug Administration. National Economic Council Director Kevin Hassett told Fox Business that the administration is studying a possible executive order to give a clear roadmap for how future AI models that create vulnerabilities should go through a process so that they’re released into the wild after they’ve been proven safe, just like an FDA drug. Bloomberg

The analogy is rhetorically clean. It is also, on closer inspection, strained in ways that matter for how any eventual mandatory regime would function in practice.

Drug approval is predicated on a relatively bounded hypothesis: does this compound do what it claims, without causing specified harms? The FDA’s clinical trial infrastructure, built over decades, evaluates outcomes in controlled populations against defined endpoints. Frontier AI models behave differently. Their capabilities emerge non-linearly from scale, training data, and interaction patterns that no pre-deployment test suite can exhaustively simulate. A model that passes a red-teaming exercise on Tuesday may discover a novel attack vector in production by Thursday.

CAISI conducts post-deployment evaluations to track risks that emerge after launch, since AI systems often behave differently under real-world conditions — including adversarial inputs and dataset drift — than they do in controlled testing environments. This acknowledgment, buried in the operational details of how CAISI works, quietly concedes what the FDA analogy papers over: there is no clean approval moment. Safety is a continuous process, not a gate. Arnav

Still, the political logic of the FDA frame is sound. It gives the administration a vocabulary for oversight that doesn’t require it to announce a regulatory regime. “Proven safe before release” is a message that plays well. The implementation will be considerably messier.

A bipartisan group of 32 House lawmakers has written to National Cyber Director Sean Cairncross urging immediate action to confront the high volume of cyber vulnerability disclosures cropping up from advanced AI systems. The letter marks an escalation in pressure on the Trump administration to confront the risks posed by frontier AI cyber models. That kind of bipartisan pressure — rare in contemporary Washington — signals that this issue has moved beyond the usual partisan channels. Axios

Second-Order Effects: Markets, Enterprise, and the Voluntary-to-Mandatory Gradient

The agreements announced on May 5 are voluntary. That status, however, may have a shorter shelf life than the companies involved are counting on.

National Economic Council Director Hassett said it’s “really quite likely” that any testing spelled out under an executive order would ultimately extend to all AI companies. “I think Mythos is the first of them, but it’s incumbent on us to build a system,” he said. When a White House economic adviser publicly floats universal applicability, the “voluntary” characterisation begins to function more as a transitional state than a permanent arrangement. Insurance Journal

For enterprise buyers, the near-term implications are more concrete. A CAISI evaluation — particularly one conducted in a classified environment, with results shared selectively across agencies — effectively creates an informal tier of government-vetted AI systems. The companies that have signed these agreements (Google DeepMind, Microsoft, xAI, OpenAI, and Anthropic) are, not coincidentally, the same companies that supply the overwhelming majority of frontier AI infrastructure to federal agencies. A new entrant — a well-capitalised European lab, or a fast-scaling domestic startup — that hasn’t been through the CAISI process faces an implicit disadvantage in federal procurement, regardless of whether any formal mandate exists.

The market signal is already visible. Following the announcement, Microsoft’s stock was down 0.6 percent in midday trading, while Alphabet, Google’s parent company, was trending in the opposite direction — up 1.3 percent. These are small moves, and reading too much into single-session trading is unwise. But the divergence may reflect a market reading of which company has the most to gain from tighter relationships with Washington’s AI oversight apparatus. Al Jazeera

The international dimension compounds the picture. The EU’s AI Act, which came into full force in August 2025, imposes mandatory conformity assessments on high-risk AI systems. The CAISI framework, built on voluntary agreements and classified evaluations, is a fundamentally different architecture — one shaped by American deregulatory instincts even as it begins to converge toward similar outcomes. The question of mutual recognition, or regulatory fragmentation, will land on the desks of trade negotiators before the decade is out.

The Counterargument: Testing Without Teeth?

Not everyone views the CAISI expansion as a meaningful check on frontier AI risk. Critics — some within the AI safety research community, others in civil liberties organisations — have raised a set of concerns that deserve a serious hearing rather than a dismissal.

The first is structural: evaluations conducted under voluntary agreements give the evaluated parties significant influence over what the evaluators can access, how results are framed, and whether findings lead to any material consequence. The new agreements allow CAISI to evaluate new AI models and their potential impact on national security and public safety ahead of their launch, and to conduct research and testing after AI models are deployed. What the agreements do not stipulate, publicly at least, is what happens when CAISI finds something troubling. The absence of a defined enforcement mechanism isn’t a technicality — it’s the central design question. CNN

The second concern is about scope creep in the opposite direction. The agreements build upon OpenAI and Anthropic’s agreements in 2024, which were the first of this kind. Each iteration has expanded the framework’s reach without a parallel expansion of CAISI’s evaluation capacity or legal authority. If the executive order now under consideration mandates testing without addressing the resource gap Jessica Ji identified, the process risks becoming a compliance ritual rather than a genuine safety check — something labs can credential-wash without fundamentally altering their deployment timelines. The Hill

Industry groups have been supportive: Business Software Alliance Senior Vice President Aaron Cooper said that CAISI brings the necessary expertise to work with private sector partners to evaluate frontier models for safety and national security risks, and called it the right institutional home within government. Industry enthusiasm for a regulatory body is not, historically, a reliable indicator of rigorous oversight. It can equally signal confidence that the oversight will remain manageable. Nextgov.com

A Framework in Formation

The agreements signed on May 5 are neither a regulatory revolution nor a fig leaf. They are something more interesting and more ambiguous than either characterisation allows.

Washington has moved from ignoring frontier AI risk to institutionalising a mechanism for examining it — in under eighteen months, and largely under the pressure of a single model’s demonstrated capabilities. That is, by the standards of government technology policy, fast. The CAISI framework exists, it has now absorbed five of the most significant frontier labs, and it has begun to develop the institutional muscle memory that eventually becomes precedent.

What it lacks is clarity on consequences. The voluntary-to-mandatory gradient that Hassett suggested — extending CAISI-style testing to all AI companies — would represent a genuine structural shift. Whether such an order arrives, and whether it comes with enforcement mechanisms or remains aspirational, will determine whether the May 5 announcements are remembered as a turning point or a photo opportunity.

The FDA comparison is imperfect. The analogy is imprecise. But the underlying instinct — that something this powerful, moving this fast, probably shouldn’t enter the world completely unexamined — is harder to argue with every week that passes.

The question now isn’t whether Washington will test frontier AI before it ships. It’s whether the testing, when it finds something, will actually matter.

Discover more from The Economy

Subscribe to get the latest posts sent to your email.

Up Next

Five PSX Stocks Worth Owning Before the Second Quarter Ends

Don't Miss

The Price of Fiscal Concord: Inside Pakistan’s Rs500 Billion IMF-Sanctioned Tax Overhaul

Click to comment

Analysis

South Korea’s Won Slides to Its Weakest Since Lehman: Asia market impact

Published

13 hours ago

July 1, 2026

Abdul Rahman

South Korea’s won has not traded at these levels since Lehman Brothers collapsed and the world was sorting through the wreckage of its worst financial crisis in eighty years. That the currency has returned to those depths under entirely different circumstances — not a global credit event, but a sustained combination of dollar strength, political uncertainty, and structural capital outflows — makes the current episode more complex, and in some ways more concerning, than 2009.

The Numbers

On July 1, 2026, the won declined as much as 0.6 percent to 1,559.10 per dollar, following a prior session low of 1,562.20 — a level last seen in March 2009. Overseas investors sold a net 1.46 trillion won ($938 million) of stocks in the Kospi index on a single trading day, marking the eighth consecutive session of equity outflows from the Korean market.

“The dollar’s strength is such that a fresh low for the won would not be surprising,” said Moon Dawoon, an economist at Korea Investment & Securities. “If it does break through, it will be difficult to identify the next technical level, so from a qualitative perspective, the downside for the won should be kept open to around 1,600 per dollar.”

A breach of 1,600 would represent territory not visited since the 1997 Asian financial crisis — a threshold that carries both technical and psychological significance for regional currency markets.

Why the Won Is Falling

The 2026 won story is not a simple export slump. South Korea continues to run a current-account surplus — $18.70 billion in December 2025, $13.26 billion in January 2026. The fundamentals of the trade balance have not deteriorated dramatically. What has changed is the capital account.

Several forces are pulling simultaneously in the wrong direction. The US-Korea interest rate differential remains wide, making dollar-denominated assets relatively attractive to Korean investors. Structural outward investment — Korean residents and institutions consistently moving capital into foreign assets — keeps upward pressure on dollar demand. Trade friction and tariff uncertainty from the United States raise risk premia on Korean assets broadly. And geopolitical stress in the Middle East has driven a risk-off flight to dollar safety that penalises emerging market currencies disproportionately.

The IMF estimated Korea’s growth at 0.9 percent in 2025, with a projected rebound to 1.8 percent in 2026 — an improvement, but well below Korea’s historical growth trajectory. The Bank of Korea has held its base rate at 2.50 percent, balancing growth support against exchange-rate and financial stability concerns.

The Semiconductor Exposure

Korea’s currency vulnerability is amplified by its sector concentration. Samsung and SK Hynix together constitute a dominant share of the global memory chip market — and global memory chip markets are themselves being stress-tested by the AI infrastructure boom. The so-called “RAMageddon” dynamic, in which AI-fuelled demand for memory chips has sent prices soaring, has provided export revenue support. But it has also created concentration risk: a reversal in AI capex demand, which the BIS and Chinese hedge funds have been warning about, would hit Korea’s export base and currency simultaneously.

The Kospi index’s heavy weighting toward Samsung, Hyundai, and semiconductor-adjacent companies means that institutional investors who reduce technology sector exposure globally tend to sell Korean equities as a primary execution path. Eight consecutive days of outflows is the market expressing that thesis in real time.

Regulatory Response

Following an earlier episode in which the won slid to its lowest since 2009 in June 2026, South Korean authorities convened an emergency meeting between the Bank of Korea governor and financial regulators. The government announced measures including stepped-up oversight of offshore currency derivatives, boosted inspections for suspected market misconduct, and investigations into potentially illegal foreign-exchange transactions.

The won briefly rebounded following those announcements before resuming its decline in early July. The pattern is familiar in currency management: administrative measures can slow momentum but rarely reverse the underlying capital flow dynamics that are driving the move.

Regional Contagion Signals

The won’s decline on July 1 led a broader retreat in Asian currencies, reflecting the dollar’s role as the default safe haven in periods of global risk aversion. The Japanese yen simultaneously extended losses to multi-decade highs against the dollar — a different dynamic driven by the US-Japan rate differential, but contributing to a picture of simultaneous stress across the major Asian currency pairs.

Emerging market investors are monitoring whether won weakness begins translating into spillover dynamics: whether Korean retail investors rotate into crypto as a won hedge (measurable through the “kimchi premium” on Korean crypto exchanges), and whether institutional outflows from Korean equity and bond markets intensify as currency losses erode total returns for foreign holders.

A currency at 1,562 per dollar, trending toward 1,600, with eight straight days of equity outflows and a semiconductor sector exposed to an AI capex cycle that global institutions are increasingly questioning — is not a crisis yet. But it is accumulating the conditions for one.

Discover more from The Economy

Subscribe to get the latest posts sent to your email.

Analysis

Japan’s $2.3 Trillion Bet: Takaichi’s AI-Semiconductor Moonshot and the Fiscal Tightrope It Requires

Published

13 hours ago

July 1, 2026

Abdul Rahman

Japan has never been timid about industrial policy. But the plan unveiled by Prime Minister Sanae Takaichi on June 24, 2026, represents an ambition of a different magnitude: JPY 370 trillion — approximately $2.3 trillion — in combined public and private investment across 17 strategic sectors over the 14 fiscal years ending in March 2041. It is the most consequential economic growth blueprint Japan has released in a generation, and it carries risks proportionate to its scale.

The Numbers and Their Logic

The plan’s centrepiece is AI and semiconductors, which together account for JPY 101.6 trillion — nearly one-third of the total. Of that allocation, the largest share targets semiconductor manufacturing. The government projects that domestic chip sales, currently at roughly 8 trillion yen annually, will reach 40 trillion yen by fiscal 2040: a fivefold increase that would require sustained policy commitment, significant private capital mobilisation, and a structural reconfiguration of Japan’s manufacturing base.

Beyond semiconductors, the plan earmarks $65 billion specifically for AI infrastructure — data centres, power capacity, and the hardware underlying large-scale AI deployment. Vertical AI tools, built for specific industries such as healthcare, manufacturing, and logistics, receive separate priority funding alongside physical AI systems. The government projects semiconductor investment alone will generate 443 trillion yen in economic spillovers by fiscal 2040, with physical and vertical AI adding a further combined 366 trillion yen.

Additional sectors covered include defence, space development, advanced manufacturing, shipbuilding, and critical minerals — all framed as pillars of economic security in an era of intensifying geopolitical competition.

The Political Context

Takaichi became Japan’s first female prime minister in October 2025, following a decisive Liberal Democratic Party electoral victory in February 2026 that gave her government the political runway to pursue long-horizon strategies. The plan builds on prior investment commitments: since 2021, the government has channelled roughly 7.2 trillion yen into semiconductors and AI, including approximately 2.6 trillion yen in support for state-backed chip venture Rapidus.

The Nikkei 225 briefly surpassed 72,000 following the announcement — a level that reflected AI-adjacent stock enthusiasm, particularly around SoftBank and Tokyo Electron. The market signal was interpretable in two ways: confidence in the industrial vision, or exuberance about government-supported capital flows into a sector already attracting speculative premium.

The Fiscal Tightrope

The plan’s fiscal architecture is where complexity enters. According to the Japanese government’s roadmap, public funding accounts for slightly less than half of the total, with the remainder expected from private capital. Three long-term fiscal scenarios were released alongside the plan, with sharply divergent outcomes.

In the most optimistic case, the strategy delivers as intended: Japan’s debt-to-GDP ratio declines steadily even as the government contributes 10 trillion yen in real annual spending. In the two alternative scenarios, where market demand or technological uptake falls short, the ratio resumes its upward trajectory during the 2030s.

Critically, all three scenarios assume inflation stabilises at around 2 percent. They exclude the potential costs of expanded defence spending and proposed consumption-tax reductions, meaning actual fiscal pressure could significantly exceed the government’s baseline projections. Meanwhile, Japan’s superlong government bond yields have risen to multi-decade highs — a market signal that investor confidence in fiscal discipline is not fully intact, even as the Nikkei rallied.

The Bank of Japan, under Governor Kazuo Ueda, has signalled continued rate increases in response to above-target inflation and upside price risks. Deputy Governor Ryozo Himino reinforced that the BoJ expects to adjust policy in response to economic conditions and financial developments, while monitoring risks including the conflict in Iran. A government pushing expansionary fiscal policy while the central bank tightens monetary conditions is a combination that creates sovereign yield risk — precisely the kind of sovereign-financial nexus the BIS has flagged as a global vulnerability.

The Industrial Security Imperative

The plan’s framing as an economic security initiative, rather than purely a growth strategy, reflects Japan’s reading of the current geopolitical moment. Supply chain resilience, technological self-sufficiency, and domestic semiconductor capacity have become strategic imperatives for governments across the developed world in the wake of the pandemic disruptions and US-China technology competition.

Japan’s bid to quinttuple domestic chip sales by 2040 places it in direct competition with the United States’ CHIPS Act investments, the EU’s European Chips Act, and South Korea’s semiconductor cluster ambitions. The difference is that Japan is making the largest single national commitment to that competition — a bet that the country has identified the window for industrial transformation, and that the cost of missing it exceeds the fiscal risk of pursuing it.

Whether the numbers work depends on outcomes that no government roadmap can control: whether AI adoption curves justify the infrastructure being built, whether Rapidus can achieve competitive semiconductor yields, and whether private capital follows government funds at the scale the plan requires. The bet is large. The stakes are higher.

Discover more from The Economy

Subscribe to get the latest posts sent to your email.

Analysis

A 13% Surge in Billionaires, a Falling Median: The AI Boom’s Wealth Paradox

Published

13 hours ago

July 1, 2026

Abdul Rahman

The numbers are unambiguous, even if their implications remain contested. In 2025, global personal wealth rose at its fastest pace since 2017. Nearly one million new millionaires were minted. The billionaire population swelled by 13 percent. And in most of the 56 markets where the UBS Global Wealth Report tracks outcomes, median wealth — the wealth of the person sitting precisely in the middle of the distribution — actually declined.

That combination, record headline growth alongside falling typical household wealth, is the defining economic signature of the AI boom. It raises questions about the sustainability of an economic narrative built on aggregate progress.

What the UBS Report Found

The UBS Global Wealth Report 2026, released June 30 and built from data spanning 56 markets representing 92 percent of all global wealth, recorded 10.8 percent growth in personal wealth in 2025 — the fastest rate in at least three years. The millionaire population grew by 1.5 percent, adding close to one million people at a pace of roughly 2,680 per day.

More than 440,000 of those new millionaires were American — exceeding 1,200 per day — making the United States responsible for close to half of the worldwide increase. The United Kingdom added more than 43,000 new millionaires, while France, Spain, Japan, and India each added more than 30,000.

The report also counted 3,302 US dollar billionaires, an increase of 383 people, or 13.1 percent, over the prior year. Billionaire wealth grew by 25 percent on average in the year ended in April, compared with a 10.8 percent rise in average personal wealth. James Mazeau, an economist at UBS, attributed the outperformance directly to the AI boom in equity markets.

The Median Paradox

UBS chief economist Paul Donovan acknowledged to Fortune what the headline figures conceal: “There is a concentration of equity wealth into the very highest wealth and income cohorts, which means that periods of strong equity performance will widen the gap between the two.” When asset markets rise and the gains are overwhelmingly held at the top of the distribution, aggregate averages can soar while the typical household experiences stagnation or decline.

The pattern is not incidental. Software and platform businesses scale at close to zero marginal cost, meaning that when an AI-adjacent product wins, it tends to win globally — and the revenue, profit, and equity all funnel into very few hands. The World Inequality Report 2026 sharpened the point with striking precision: just 56,000 ultra-wealthy individuals — the top 0.001 percent — now control more wealth than the poorest 4 billion people on Earth combined. Their share of global wealth has nearly doubled since 1995.

Since 1995, billionaire wealth has compounded at approximately 8.5 percent annually. The bottom half of the global population has grown theirs at roughly 3.4 percent.

The Ultra-Wealthy Tier Accelerates

Altrata, a wealth intelligence firm, tracked a 14.4 percent jump in 2025 in the number of people worth more than $30 million — reaching a record 556,850 worldwide. In mainland China, the $50 million to $100 million cohort has compounded in real terms at nearly 31 percent annually since 2000. The United States’ top 1 percent of households, per the Federal Reserve, now holds approximately 32 percent of the nation’s total wealth — the highest proportion since the Fed began compiling the relevant data in 1989.

Within this hierarchy, the AI trade has functioned as a supercharger. Founders who hold large equity stakes in companies that have benefited from AI-driven market re-ratings have watched their personal wealth compound at the same exponential rates as the underlying businesses. The upcoming major IPOs — SpaceX, Anthropic, and OpenAI — are projected to create a new cohort of billionaires and dramatically expand the existing ultrawealthy population.

The Political Economy of the K-Shape

Bloomberg’s K-shaped economy analysis projected that the divergence between asset holders and wage earners will deepen further. The political consequences are already visible. California Governor Gavin Newsom, in comments reported ahead of a potential 2028 presidential run, proposed a national wealth tax and an initiative to give Americans a direct stake in AI development. Former Amazon CEO Jeff Bezos called for the bottom 50 percent of earners to pay zero federal income tax.

Axios reported that a growing number of tech billionaires are developing prescriptions for AI-fuelled inequality — not from altruism, but from a calculation that populist revolt represents a greater threat to their interests than redistributive taxation. “The pitchforks are here, they’re not just coming,” Newsom warned, predicting that resentment toward billionaires and AI-driven automation will dominate the 2026 and 2028 electoral cycles.

Donovan, the UBS economist, noted that governments are likely to seek to mobilise wealth to lower the cost of debt finance. What that means in practice — wealth taxes, forced investment mandates, or some novel fiscal instrument — remains the defining policy question of the decade the AI boom is creating.