The screen bled red across the trading floors of Lower Manhattan on Tuesday, pulling the curtain down on a euphoric 18-month rally. As the closing bell rang, a brutal Nasdaq AI stock sell-off had wiped out 3% of the index’s value, vaporising hundreds of billions in market capitalisation in mere hours. Yet, step away from the glare of the tech titans, and the picture shifts entirely. Small-cap industrials, regional banks, and consumer staples quietly advanced. This was not a panic. It was a surgical, deeply concentrated liquidation event targeting the very silicon and software giants that have single-handedly dragged global markets to record highs.

To understand the severity of this capital rotation, one must look at the immense concentration risk that preceded it. By late May, just five artificial intelligence bellwethers accounted for roughly 30% of the S&P 500’s total market weighting. This is a historical anomaly surpassing even the dot-com peak of early 2000. Institutional portfolios had become dangerously top-heavy. When momentum cracked, the reversal was violent.

Data from financial market trackers at Reuters revealed that trading volumes for semiconductor equities surged 45% above their 30-day moving average during the afternoon session. This mass exit eclipsed the broader market’s reality. According to global market analysis from Bloomberg, the S&P 500 equal-weight index actually closed in positive territory, highlighting a stark bifurcation. Investors aren’t fleeing equities; they’ve simply decided to cash out their AI lottery tickets and move funds into the forgotten corners of the real economy.

The mechanics of a Nasdaq AI stock sell-off rarely start with a scream; they start with a whisper in the options market. On Monday evening, institutional hedging activity spiked, signalling that major funds were quietly locking in profits on their semiconductor and cloud computing holdings. By Tuesday morning, that defensive posturing erupted into outright selling.

The trigger was a combination of stretched valuations and exhaustion. Nvidia, which had priced in a near-perfect trajectory of endless exponential growth, saw its forward price-to-earnings multiple rejected by the market. When shares of the chipmaker plunged, it dragged the entire semiconductor index down with it. A market analysis brief from the Financial Times noted that almost $400 billion in semiconductor market capitalisation evaporated in the first 90 minutes of trading alone.

That is roughly equivalent to the entire GDP of Denmark vanishing before lunch.

Still, the destruction was highly selective. Software-as-a-service providers that had recently slapped artificial intelligence onto their investor decks without demonstrating corresponding revenue growth faced the harshest penalties. Valuations in this speculative tier contracted by double digits. The market is abruptly demanding proof of concept. Generative models are expensive to train, and Wall Street won’t fund the capital expenditure without a clear line of sight to immediate profitability.

Analysts at the International Monetary Fund recently warned of this exact vulnerability, calculating that tech sector multiples had become unmoored from historical norms, leaving them acutely exposed to sudden sentiment shifts. When the narrative changed, the algorithmic trading desks amplified the slide, triggering a cascade of automated stop-loss orders. Yet, the devastation was quarantined. Outside the tech-heavy indexes, the Dow Jones Industrial Average held steady, buoyed by traditional blue-chip stocks. This divergence reveals a market that isn’t experiencing a macro-economic failure, but rather a violent recalibration of pricing in its most overextended sector.

Why a Tech Sector Correction Was Inevitable

To view Tuesday’s rout as a sudden shock is to ignore months of flashing warning lights. The market had entered a phase of inelastic exuberance. Every mention of machine learning by a Chief Executive on an earnings call was met with a blind surge in share price, creating a dangerous feedback loop of capital misallocation. The fundamental laws of financial physics were suspended, but only temporarily.

Why are AI stocks dropping? They are falling because investors have realised that the timeline for artificial intelligence to generate enterprise-level profits is vastly longer than the timeline required to build the infrastructure. Valuations priced in immediate perfection, leaving no margin for delayed adoption, regulatory hurdles, or rising capital expenditure costs.

This tech sector correction is a symptom of market digestion. The “Magnificent Seven” and their supply chains had absorbed nearly all available retail and institutional liquidity over the past year. But as the third quarter approaches, the burden of proof is shifting. Companies are now expected to demonstrate exactly how their massive investments in graphics processing units translate into bottom-line free cash flow. For many, the math simply doesn’t add up yet.

That said, the rotation out of these names is structurally healthy. When capital pools exclusively in one sector, it starves the rest of the market of investment. The fact that capital is flowing from overvalued tech darlings into energy, materials, and healthcare suggests that the underlying economy remains resilient, even if the speculative edge has been blunted. The current semiconductor stock drop is stripping the froth from the market, punishing tourists who bought the ticker symbol rather than the balance sheet. We are witnessing a transition from a momentum-driven market to one that prioritises earnings quality. The era of the blank cheque has officially closed.

The downstream consequences of this capital rotation will reshape venture capital, corporate strategy, and perhaps even monetary policy over the next 12 months. The immediate victim will be the private markets. Startup founders who have spent the last year riding the coattails of public market valuations will face a brutal awakening. Seed funding rounds that previously commanded astronomical valuations based on a sleek demo will now face rigorous due diligence. The hurdle rate for new capital just went up.

For corporate boards, the message is equally stark. The market will no longer reward performative spending. Executives who have engaged in an arms race to acquire compute power will now be pressured by activist investors to justify those expenditures. If the infrastructure doesn’t yield margin expansion or significant productivity gains, those tech budgets will be slashed. This creates a secondary risk for the chip designers and cloud providers: their current revenue run-rates are highly dependent on this very corporate arms race. If enterprise spending slows, the revenue models of the tech giants will need to be drastically revised.

From a macroeconomic perspective, this deflation of the AI market bubble may actually provide the Federal Reserve with a measure of comfort. According to research published by the World Bank, hyper-concentrated equity rallies can create artificial wealth effects that complicate inflation targeting. By cooling off the most speculative corners of the market, the central bank may find it easier to manage the broader economic glide path without triggering a deep recession. The destruction of paper wealth in Silicon Valley doesn’t immediately translate to job losses on Main Street. Instead, the normalisation of a Nasdaq 100 decline removes a significant source of systemic risk. The coming quarters will be defined by an intense focus on margins, operational efficiency, and the arduous task of turning a dazzling science project into a viable corporate utility.

What follows, however, is fiercely debated. Not everyone interprets this sell-off as a return to fundamental sanity. A vocal contingent of market strategists argues that abandoning the trade now is akin to selling internet infrastructure stocks in 1998 — a premature exit from a generational wealth-creation cycle.

Their argument rests on the sheer scale of the technological shift. Generative models aren’t merely a new software vertical; they are a general-purpose technology comparable to the internal combustion engine or electricity. A recent analysis by the OECD points out that artificial intelligence integration could increase global labour productivity by up to 1.5 percentage points annually over the next decade. If that thesis holds true, the current valuations of the top silicon producers and cloud hyper-scalers are actually conservative, not stretched.

From this perspective, Tuesday’s decline is nothing more than a momentary blip. It is viewed as a liquidity-driven shakeout designed to clear weak hands from the market. The bulls argue that the massive capital expenditures by the tech giants aren’t a sign of excess, but a necessary moat-building exercise. They contend that the broader market is overestimating the risk of delayed adoption and underestimating the exponential curve of computing power. If they are right, the capital rotating into defensive stocks today will eventually be forced back into the tech sector at a severe premium, missing the next massive leg of the rally.

The tension between these two realities — the undeniable long-term transformative power of machine learning and the immediate, punishing math of overextended equity valuations — will dictate market dynamics for the foreseeable future. Tuesday’s brutal correction was not an indictment of the technology itself, but a rejection of the timeline investors had assigned to it. The market is demanding a return to financial gravity. Capital hasn’t evaporated; it has simply grown impatient, seeking refuge in the unglamorous, cash-generating sectors of the old economy while the new economy figures out its business model.

The AI revolution is far from over, but the easy money has already been made.

Northern Trust’s $1.4 trillion asset management arm says the AI boom is “massively disinflationary.” The evidence is building — but so are the near-term headwinds. Here is what the bulls are getting right, what they are glossing over, and what every central banker should be thinking about this week.

Analysis · 2,150 words · Cites: Northern Trust, IMF WEO April 2026, BIS Working Papers, OECD

There is a sentence making the rounds in macro circles this morning that deserves more than a tweet. Northern Trust Asset Management — custodian of $1.4 trillion in client assets — told the Financial Times that the AI boom is poised to be “massively disinflationary.” Two words, twelve letters, and an argument that, if it proves correct, will reshape monetary policy for the rest of this decade. If it proves wrong, it will look like the most expensive case of group-think in asset management history.

The claim is bold, but it is not baseless. Across its 2026 Capital Market Assumptions, Northern Trust has laid the groundwork: nearly 40 percent of jobs worldwide — and 60 percent in advanced economies — are now exposed to AI, signalling what the firm calls “a major shift” in productivity and labor market dynamics. Add to that the IMF’s own January 2026 estimate that rapid AI adoption could lift global growth by as much as 0.3 percentage points this year alone, and up to 0.8 percentage points annually in the medium term, and suddenly “massively disinflationary” sounds less like a marketing line and more like a macroeconomic thesis worth taking seriously.

But serious theses deserve serious scrutiny. And when you peel back the optimism, you find a story with a considerably more complicated second act.

“AI today is still in its early innings. It is reshaping how we operate. It is reshaping how we work. Yet at the same time, we know there are going to be a number of missteps.” — Northern Trust Asset Management, February 2026

The disinflationary logic — and why it is compelling

The core argument runs as follows. AI raises the productive capacity of every worker, firm, and economy that adopts it. More output from the same inputs means falling unit costs. Falling unit costs mean downward pressure on prices. In a world still wrestling with inflation — the IMF’s April 2026 World Economic Outlook projects global headline inflation at 4.4 percent this year, elevated partly by a new Middle East conflict — that kind of structural supply-side boost could not arrive at a better moment.

The historical analogy is not perfect, but it is instructive. The internet and personal computing drove a productivity renaissance through the 1990s that helped the US run a decade of growth with unusually low inflation. The difference this time, optimists argue, is both speed and scope. Generative AI is being deployed across sectors — finance, law, medicine, logistics, software — simultaneously, rather than trickling through the economy over fifteen years. The IMF’s own research noted that investment in information-processing equipment and software grew 16.5 percent year-on-year in the third quarter of 2025 in the United States alone. That is not a technology cycle. That is a structural reorientation.

At the firm level, the mechanism is equally legible. AI-assisted coding reduces software development costs. AI-powered customer service reduces headcount requirements per unit of output. AI-accelerated drug discovery compresses R&D timelines. Each of these reduces costs for producers, and in competitive markets, cost reductions eventually become price reductions for consumers. The BIS, in its 2026 working paper on AI adoption among European firms, found measurable productivity gains at companies with higher AI adoption rates — gains that, if broad-based, translate directly into disinflationary pressure.

The uncomfortable counterarguments

Now for the cold water. The hyperscalers — Alphabet, Microsoft, Amazon, Meta — are expected to spend upwards of $600 billion on data center capital expenditure in 2026 alone, according to Northern Trust’s own analysis. That is $600 billion of demand competing for semiconductors, specialised labor, land, electricity infrastructure, and cooling systems. In the near term, this is not disinflationary. It is, by any honest accounting, inflationary. It bids up the price of every input that AI infrastructure requires.

Energy is the most acute example. Northern Trust’s own economists have noted that data centers are expected to account for 20 percent of the increase in global electricity usage through 2030. The IMF’s recent research put it plainly: energy bottlenecks “could delay AI diffusion, anchor a higher level of core inflation, and generate local pricing pressures” in grid-constrained regions. This is not a theoretical risk. It is a live constraint in the US, the UK, Ireland, Singapore, and across northern Europe, where grid capacity has become a hard ceiling on data center expansion.

There is also the measurement problem — and it is a serious one. As the IMF’s own Finance & Development noted in its March 2026 issue, GDP accounting simultaneously overstates AI’s immediate contribution (by counting massive capital outlays as output) while understating its broader economic impact (by missing productivity spillovers that do not show up in standard national accounts). This is precisely the statistical paradox that masked the early productivity gains of the 1990s IT revolution — and it cuts in both directions for policymakers. If AI is quietly raising potential output, the economy may be running cooler than headline data implies. If the infrastructure surge is instead stoking a new floor for energy and construction costs, central banks may be tightening into a real supply shock.

The IMF’s chief economist Pierre-Olivier Gourinchas put the dilemma with characteristic precision: the AI boom could lift global growth, but it also “poses risks for heightened inflation if it continues at its breakneck pace.” That is the paradox in miniature — the same technology that promises to lower prices over time is currently consuming enormous resources to build itself.

The geopolitical dimension: who wins, who lags, and who is locked out

The disinflationary thesis is not uniformly distributed across the global economy, and this is where the Northern Trust framing risks glossing over structural inequality. Advanced economies — the US, Japan, Australia, South Korea — are positioned to capture the productivity upside first. Their firms are adopting, their labor markets are adapting, and their capital markets are pricing in the gains. Northern Trust’s own forecasts identify the US, Japan, and Australia as likely leaders in equity returns over the next decade, precisely because of AI-driven productivity.

Europe sits in a more ambiguous position. The continent is not at the forefront of AI model development, and Northern Trust acknowledges it explicitly in its CMA 2026. The region offers a healthy dividend yield and attractive valuations — but if AI productivity is the driver of the next decade’s returns, Europe’s relative lag in AI infrastructure and frontier model development is a structural disadvantage, not a cyclical one. The ECB faces its own version of the monetary policy puzzle: if AI-driven disinflation arrives later and slower in Europe than in the US, it changes the rate path, the currency dynamics, and the comparative fiscal math.

Emerging markets face the starkest challenge. The IMF’s analysis of AI in developing economies is clear: AI preparedness — digital infrastructure, human capital, institutional capacity — is the binding constraint on whether productivity gains materialize or get captured entirely by technology importers. Many emerging economies are primarily consumers of AI built elsewhere. The disinflationary benefits they receive are mediated through imports; the inflationary effects of AI-driven energy demand and semiconductor scarcity are borne locally. The net result, without deliberate policy intervention, is a widening productivity gap rather than a convergence story.

China deserves a separate paragraph. Its AI investment is substantial and accelerating, even under the constraints of US semiconductor export controls. The China-US AI race is not merely a geopolitical contest — it is a race to determine which economy gets to define and monetize the next general-purpose technology. Beijing’s capacity to deploy AI at scale across manufacturing, logistics, and services could generate its own disinflationary dynamic, although its ability to export that technology — and the disinflation it carries — is constrained by the very geopolitical tensions that are simultaneously driving energy and defence inflation.

What central banks should actually do

The honest answer is: proceed carefully, communicate transparently, and resist the temptation to read AI’s structural effect through the noise of its near-term capex cycle. The IMF’s April 2026 World Economic Outlook makes the right call when it urges central banks to guard against “prolonged supply shocks destabilising inflation expectations” while reserving the right to “look through negative supply shocks” where inflation expectations remain anchored.

That is the narrow path. If AI is genuinely raising potential output, then central banks that tighten aggressively in response to near-term energy and infrastructure inflation are making a classic policy error: fighting tomorrow’s economy with yesterday’s models. The 1990s analogy is instructive again — the Federal Reserve’s willingness to allow growth to run above conventional estimates of potential, on the grounds that productivity was accelerating, helped produce the longest peacetime expansion in American history.

But the reverse error is equally dangerous. If the AI productivity jackpot takes longer to arrive than Northern Trust and its peers anticipate — and Daron Acemoglu’s careful 2025 work in Economic Policy gives serious reason for that caution — then central banks that ease prematurely, trusting in a disinflationary future that is still several years away, risk entrenching the very inflation they spent the early 2020s battling back.

The IMF is right to treat AI as what it called in its April 2026 research note “a macro-critical transition rather than a standard technology shock.” Human decisions — by managers, workers, regulators, and investors — will shape the pace of adoption, the distribution of gains, and the political sustainability of the disruption. Those decisions are not made yet. Which means the data, for now, is genuinely ambiguous.

The verdict: right thesis, wrong timeline

Northern Trust is probably correct that AI will be massively disinflationary. The logic is sound, the historical analogies are supportive, and the scale of investment being made is simply too large to yield no productivity dividend. The question is not whether, but when — and the “when” matters enormously for portfolio construction, monetary policy, and fiscal planning.

The near-term picture, stripped of AI optimism, is one of elevated global inflation shaped by geopolitical conflict, persistent services price stickiness, and a capex boom that is consuming rather than producing cheap goods. The medium-term picture, contingent on adoption rates and diffusion across the global economy, is one where AI-driven productivity could deliver a genuine and sustained disinflationary impulse — the kind that would allow central banks to run looser for longer, equity multiples to expand sustainably, and real wages to recover.

The investor who misidentifies the timeline — and treats the medium-term story as immediate reality — will find themselves long duration in a world where rates stay higher than expected, and long AI infrastructure capex in a world where the ROI question remains, as Northern Trust itself acknowledged in February, one of “many more questions than answers.”

The honest macro position, as of April 2026, is this: Northern Trust is pointing in the right direction. But they may be holding the map upside down with respect to the calendar. For investors, policymakers, and strategists, the discipline required is not deciding whether AI will be disinflationary — it will — but calibrating, with intellectual humility, exactly how long the world will have to wait before the price deflator actually arrives.

The most consequential financial-security meeting of 2026 happened Tuesday. Almost nobody was talking about it.

There is a particular quality to urgency in Washington — a calibrated, deliberate kind, stripped of drama precisely because the stakes are too high for theater. When Treasury Secretary Scott Bessent and Federal Reserve Chair Jerome Powell jointly summon the chiefs of America’s largest banks to a private session on a weekday morning, they are not performing concern. They are managing it.

That is what happened on Tuesday, April 8, 2026, in the marbled corridors of Treasury headquarters on Pennsylvania Avenue. Bessent and Powell assembled a group of Wall Street leaders to make sure banks are aware of possible future risks raised by Anthropic’s Mythos model and potential similar systems, and are taking precautions to defend their systems. Bloomberg The CEOs of Citigroup, Morgan Stanley, Bank of America, Wells Fargo, and Goldman Sachs were present. JPMorgan’s Jamie Dimon was invited but unable to attend. AOL The Treasury declined to comment. The Fed declined to comment. Anthropic had no immediate comment.

In Washington, silence of that particular texture is its own form of communication.

The Model That Spooked the Regulators

To understand why two of America’s most powerful financial stewards convened an emergency summit with the chiefs of institutions collectively managing trillions in assets, you need to understand what Anthropic’s Claude Mythos Preview actually does — and why it is genuinely different from the parade of large language models that have cycled through headlines since 2022.

Anthropic launched the powerful Mythos model earlier this week but stopped short of a broad release, citing concerns it could expose previously unknown cybersecurity vulnerabilities. The company said the model is capable of identifying and exploiting weaknesses across “every major operating system and every major web browser.” RTÉ Read that sentence again. Every major operating system. Every major web browser. This is not a chatbot that occasionally hallucinates. This is an autonomous vulnerability-hunting engine with the precision of an elite red team and the speed of software.

Unlike typical consumer-facing AI tools, Mythos is geared toward cybersecurity software engineering tasks. Its specialty is identifying critical software vulnerabilities and bugs, but it can also assemble sophisticated exploits. CoinDesk The distinction matters enormously. Most AI models are generative — they produce text, images, code. Mythos is analytical and adversarial, capable of scanning codebases, identifying failure points invisible to human auditors, and constructing the exploits that could weaponize those failures. In the hands of a sophisticated actor — a state-sponsored hacking collective, a ransomware syndicate, a rogue insider — this capability is not a cybersecurity tool. It is a cybersecurity threat.

This marked the first time Anthropic had limited the launch of a new model. Investing.com That fact alone should arrest attention. A company whose business model depends on broad adoption and API revenue made the deliberate, commercially costly decision to gate access. That restraint — unusual in a sector that tends to race toward release — signals something about how seriously Anthropic’s own researchers regard what they have built.

Project Glasswing: An Experiment in Controlled Power

Access to Mythos will be limited to about 40 technology companies, including Microsoft and Google, and Anthropic has been in ongoing talks with the U.S. government about the model’s capabilities. AOL This restricted release program, referred to internally as Project Glasswing, is a deliberate inversion of how AI has historically been deployed: rather than releasing broadly and patching later, Anthropic gave dominant platform holders a head start — not to monetize first, but to defend first. Anthropic released the model to a select group of partners, including Amazon, Apple, and Microsoft, to give them a head start on securing vulnerabilities. Investing.com

It is a genuinely novel approach, and one that deserves more credit than it will likely receive. The logic is sound: if a model can identify zero-day vulnerabilities at machine speed, the most responsible action is to arm defenders before the broader landscape of threat actors can replicate or steal the capability. But Glasswing also exposes a governance gap so wide you could park an aircraft carrier in it.

Who audits the 40 companies with access? What safeguards prevent Mythos from being fine-tuned, transferred, or reverse-engineered? If a Glasswing participant suffers a breach — and given that these are themselves high-value targets, the probability is non-trivial — what is the liability chain? What is the protocol? The answers to these questions do not exist in any regulatory framework currently operative in the United States, the European Union, or anywhere else.

The Systemic Risk Nobody Has Priced

The meeting at Treasury was not primarily about Anthropic. It was about what Anthropic represents: the arrival of AI capabilities that move faster than the regulatory, legal, and institutional machinery designed to contain them.

Consider the financial system’s exposure. Modern banking infrastructure is built on decades of accumulated code — legacy COBOL systems at regional lenders, middleware connecting trading platforms to clearing houses, authentication layers protecting retail deposits. Much of this code has never been audited by a sophisticated adversary because auditing at scale was prohibitively expensive. Mythos eliminates that constraint. A well-resourced actor with access to comparable capability could, in principle, systematically map the attack surface of an entire national banking system in the time it currently takes a human security team to review a single subsystem.

The episode highlights a fundamental change in how regulators are framing AI risk — not merely as a technological challenge, but as a potential catalyst for systemic events. This has already raised red flags in crypto, where experts are worried that Mythos’ capability of discovering and exploiting zero-day vulnerabilities in real-time at low cost poses risk to the DeFi infrastructure. CoinDesk

The systemic risk framing is the right one — and it is the framing that explains why Powell was in that room. The Federal Reserve’s mandate is financial stability. Historically, stability threats have come from credit cycles, liquidity crunches, and contagion. They are now coming from code. A successful AI-enabled attack on a major custodial bank — one that compromised transaction integrity, corrupted ledger data, or triggered a cascade of failed settlement — would represent a category of financial crisis that no existing playbook addresses. The bazooka of emergency liquidity provision is not particularly useful when the crisis is epistemic rather than financial: when the question is not whether there is enough money, but whether the numbers can be trusted at all.

Anthropic vs. the Pentagon: The Contradiction at the Heart of AI Policy

There is a peculiar irony shadowing this episode. Anthropic has separately been battling the Trump administration in court. The Pentagon had labeled the company as a supply-chain risk, a designation that Anthropic has opposed. Earlier this week, a federal appeals court declined, at least for now, Anthropic’s request that it put a pause to the Pentagon’s designation. Bloomberg Law

Anthropic proactively briefed senior U.S. government officials and key industry stakeholders on Mythos’s capabilities RTÉ — engaging responsibly with the national security community — even as one branch of that same government has labeled the company a security liability. The left hand of the U.S. government calls in Anthropic’s most advanced model to warn bankers about cyber risk; the right hand designates its maker a supply-chain threat. This is not incoherence. It is the natural consequence of applying 20th-century institutional categories to 21st-century technology companies that are simultaneously strategic assets, potential vulnerabilities, and independent actors with their own governance philosophies.

The contradiction will not resolve itself. It requires a policy architecture that does not currently exist — one that can hold together the dual realities that Anthropic’s capabilities are a genuine national asset and that Anthropic’s capabilities require genuine national oversight. Neither a blanket clearance nor a blanket designation captures that complexity.

What Bessent and Powell Actually Did — and What It Implies

The convening itself was a significant signal. Bessent and Powell do not share a conference room casually. The joint appearance invested the meeting with the authority of both fiscal and monetary sovereign — the message being that AI cyber risk is no longer a niche technology-sector concern but a macro-prudential one. Banks should be pricing this into their operational risk frameworks. Insurers will follow. Rating agencies will not be far behind.

But signals, however weighty, are not architecture. The meeting produced no public guidance, no regulatory proposal, no framework for how banks should report, manage, or disclose AI-enabled cyber exposures. The CEOs who left Treasury on Tuesday left with warnings — and no rulebook.

The Governance Gap and How to Begin Closing It

The Mythos episode crystallizes three failures that policymakers now have no excuse for ignoring.

First, the pre-release consultation gap. Anthropic did the right thing in briefing U.S. officials before releasing Mythos. But that consultation was informal, voluntary, and ad hoc. The EU AI Act’s tiered risk framework is imperfect, but it at least establishes mandatory pre-market assessment for high-risk systems. The United States has no equivalent. A model capable of autonomously discovering and exploiting zero-days across every major OS and browser is, by any reasonable definition, a high-risk system. Its release should trigger a formal, structured national security review — not a phone call.

Second, the systemic-risk classification vacuum. The Fed can designate non-bank financial institutions as systemically important. It cannot currently designate AI models as systemically risky. That gap is now visible and consequential. What is needed is not a new agency but a clear cross-agency mandate — Treasury, CISA, the Fed, the OCC — with authority to classify certain AI capabilities as requiring coordinated disclosure, pre-release review, and sector-specific defensive preparation.

Third, the liability architecture. If a bank suffers losses traceable to an AI-enabled attack using capabilities derived from or analogous to a commercially released model, who bears what responsibility? The current answer — whatever tort law eventually produces — is wholly inadequate for systemic risks. Liability frameworks that can price and allocate AI-era cyber risk are not a luxury. They are a precondition for insurability and, ultimately, for financial stability.

A New Era of Risk — and Responsibility

There is a version of this story that ends badly: a race between capability development and governance in which capability wins by a decisive margin, and the first major AI-enabled financial system attack comes before any of the above frameworks exist. That version is not inevitable, but it requires active work to prevent.

The Tuesday meeting at Treasury was, in its way, a hopeful sign. It suggests that the United States’ most senior financial authorities understand, at least viscerally, that the risk is real and that the clock is running. It suggests that some version of public-private coordination is possible, even in a regulatory environment that remains deeply fragmented.

Anthropic has previously disclosed that it consulted with U.S. officials ahead of Mythos’ release regarding both its defensive and offensive cyber capabilities. CoinDesk That consultation should become a standard, not an anomaly. The release of any AI system with demonstrated offensive cyber capabilities — the ability to identify and exploit zero-days at scale — should automatically trigger a mandatory interagency review, sectoral briefings for affected industries, and a public risk disclosure, however carefully worded.

What Bessent and Powell did on Tuesday was, in the truest sense, firefighting. The fire is real. But what the financial system needs is not better firefighters. It needs buildings that are harder to burn.

The Mythos moment is a clarifying one. It tells us, with unusual precision, that the era of AI as a productivity story is over. The era of AI as a security story — a national security story, a financial security story, a systemic stability story — has arrived. Policymakers who treat it otherwise are not being optimistic. They are being negligent.