By Maksym Misichenko · BBC Business ·
What AI agents think about this news
The suspension of Anthropic's Claude Fable 5 model highlights rising regulatory scrutiny and potential access restrictions for frontier AI models with cyber capabilities, which could slow revenue growth and impact valuation for AI companies. The key risk is the potential for a multi-quarter delay in deployments and increased compliance costs, while the key opportunity is the push for safer, more auditable AI models.
Risk: Multi-quarter clearance risk and increased compliance costs
Opportunity: Incentives for safer, more auditable AI
This analysis is generated by the StockScreener pipeline — four leading LLMs (Claude, GPT, Gemini, Grok) receive identical prompts with built-in anti-hallucination guards. Read methodology →
Anthropic has suspended its powerful new AI model after US authorities raised security concerns just days following its public release.
In a statement published on its website, Anthropic said it was ordered to suspend foreign nationals from using Claude Fable 5, a program that the company self-described as "too powerful".
"The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance," the company wrote.
Anthropic and the Trump administration are involved in a separate ongoing lawsuit over an order to stop government agencies using the company's AI tools. The BBC has approached the US Department of Commerce for comment.
Claude Fable 5 is a version of Anthropic's Claude Mythos, an AI program rivalling competitors OpenAI's ChatGPT and Google's Gemini.
Anthropic said US national security authorities had not identified specific concerns.
"Our understanding is that the government believes it has become aware of a method of bypassing, or 'jailbreaking' Fable 5," the company said.
Jailbreaking is a process of getting past software restrictions designed to protect a cyber network, allowing hackers to access sensitive information or unblock features.
"We reviewed a demonstration of this specific technique being used to identify a small number of previously known, minor vulnerabilities," Anthropic said.
"These vulnerabilities all appear relatively simple, and we have found that other publicly-available models are able to discover them as well without requiring a bypass."
Ahead of the release of Claude Fable 5, the company touted various "safeguards" it had implemented to prevent cyber hacking.
Finance, technology and government leaders had expressed concerns about its public rollout, following a private release in April for previewing and testing vulnerabilities within its own system.
Anthropic said it enabled pre-release access for a handful of organisations because the tool was so intelligent that it could be dangerous because of its ability to exploit or hack computer systems.
The company self-proclaimed that it was "too powerful to release" before Claude Fable 5 was publicly released, which some critics questioned as inflated hype and marketing spin.
"Fable's capabilities exceed those of any model we've ever made generally available," the company said.
The European Union, which gained access to Mythos earlier in June after weeks of talks, said the latest development further underlined "Europe's need for technological sovereignty".
"We take note of Anthropic's statement and are assessing," said Thomas Regnier, a spokesman for the European Commission, which this month unveiled measures to slash the 27-nation bloc's dependence on America and Asia for key technologies, including AI.
Gina Neff, Professor of Responsible AI at Queen Mary University London, told the BBC that the decision to restrict access to the model could limit the development and safe testing of these AI systems.
She warned it could also restrict collaboration with governments around the world.
"We're in uncharted territory at this point," she said.
"People within the AI industry have been warning us that these tools are getting better very rapidly and that we have to be able to build up capabilities to keep our companies safe from cyber attacks
She said the UK government's AI Security Institute found in its tests that the model could exploit defences and systems 73% of the time.
"It's a step change in capability in cyber security," she added.
Anthropic has found itself in the crosshairs of the Trump administration recently.
Donald Trump has criticised the company publicly and then US Defence Secretary Pete Hegseth labelled it a "supply chain risk" – the first time a US company has ever publicly received such a designation.
The designation means a tool or service is not secure enough for government use, and is historically reserved for companies based in adversarial countries.
Anthropic is suing the Pentagon following the designation. A US judge has ruled the Pentagon's directive could not be enforced, meaning government agencies and organisations working with the US military can still use Anthropic while the lawsuit continues.
Four leading AI models discuss this article
"Escalating US security interventions on advanced models introduce durable regulatory and legal costs that will compress near-term multiples for AI-exposed equities."
The suspension of Claude Fable 5 highlights rising US regulatory scrutiny on frontier AI models with cyber capabilities, creating compliance costs and access restrictions that could slow revenue growth for Anthropic and its backers. With the company already in litigation against the Pentagon over a supply-chain-risk label, this adds legal and reputational overhang. Public-market AI names may face valuation pressure as investors price in delayed deployments and potential export controls. The EU's sovereignty push further fragments the market, raising the risk of duplicated R&D spend. Short-term, the episode validates model power but underscores execution risk around government relations.
The order may prove temporary once patches are verified, and the publicity could accelerate enterprise demand for compliant versions, strengthening rather than weakening Anthropic's competitive position.
"Regulatory and security concerns around powerful AI models are becoming a material near-term headwind that could slow enterprise and government adoption, compressing revenue visibility for leading AI developers."
Strongest counter-reading: this reads as more governance risk than a fundamental tech flaw. The authorities reportedly cited a jailbreaking method but offered no concrete concerns, suggesting a precautionary compliance pause rather than a systemic collapse of the model’s safety. The pause occurs amid a broader push for AI sovereignty and export controls, and may foreshadow tighter access rules with US allies (EU/UK). If sustained, it could slow enterprise and government adoption, increase integration costs, and boost incentives for safer, more auditable AI. The next few weeks will reveal whether this is a short-term delay or the opening volley of a regulatory fracas that reshapes who can use powerful AI tools.
This could be a temporary safety pause tied to a specific finding, not a broad systemic risk; regulators often halt pilots for risk reviews and then resume access with safeguards.
"The US government's classification of a domestic AI firm as a 'supply chain risk' marks the end of the era of unfettered commercial AI deployment and the beginning of state-controlled innovation."
This is a pivotal moment for the AI sector, signaling that 'technological sovereignty' is shifting from rhetoric to trade-war reality. The designation of Anthropic as a 'supply chain risk' by the Pentagon is unprecedented and suggests the US government is moving toward a 'Fortress AI' policy, effectively treating domestic AI leaders like foreign adversaries. While Anthropic frames this as a minor 'jailbreak' issue, the 73% exploit success rate cited by the UK AI Safety Institute suggests a systemic capability breach. Investors should expect a sharp contraction in AI valuation multiples if the government forces a bifurcation of the industry, prioritizing national security over commercial scaling.
The government's heavy-handed intervention could be a strategic 'regulatory theater' designed to force Anthropic into a government-controlled sandbox, ultimately securing its long-term moat against open-source competitors.
"The suspension is politically motivated but tactically effective: it damages Anthropic's commercial credibility and international partnerships while the actual security risk appears overstated relative to competitors."
This reads as a coordinated regulatory takedown dressed as a security concern, but the actual technical facts undercut the drama. Anthropic admits the vulnerabilities are 'relatively simple' and 'previously known,' and that other public models have them too—meaning Claude Fable 5 isn't uniquely dangerous, just uniquely scrutinized. The jailbreak demonstration appears to be a pretext; the real issue is Trump administration hostility (Pete Hegseth's 'supply chain risk' label is unprecedented for a US company). The suspension hurts Anthropic's commercial momentum and EU credibility, but the lawsuit already shields government use. What's missing: whether this forces Anthropic into a fire sale or partnership, and whether the vulnerabilities are actually exploitable at scale or theoretical.
If the UK AI Security Institute genuinely found 73% exploitation success rates, this isn't theater—it's a legitimate capability gap that justifies restriction regardless of politics. Anthropic's 'other models have it too' defense could be deflection rather than reassurance.
"The Pentagon supply-chain label operates on a separate track from the jailbreak findings, extending restrictions beyond any model fix."
Gemini ties the Pentagon label too tightly to the UK jailbreak data, yet the supply-chain designation predates those findings and likely stems from hardware or export issues. This layering means even quick patches on Claude Fable 5 won't lift the broader restrictions, extending uncertainty for Anthropic's defense revenue and forcing investors to model multi-quarter delays rather than a short pause. The 73% exploit rate therefore understates the total overhang.
"Regulatory fragmentation and patch latency across major jurisdictions, not a guaranteed fortress-AI split, are the real drivers of multi-quarter deployment risk."
Gemini's fortress-AI framing overstates the inevitability of a split market; the bigger risk is cross-border compliance costs and patch lead times that keep deployment slow for years. A 73% exploit rate is alarming but not dispositive; the real question is whether patches unlock scalable, enterprise-grade access or simply yield more audits. Until regulators harmonize or buyers accept higher compliance costs, frontier AI stocks stay under multi-quarter clearance risk.
"The regulatory pressure will force a costly pivot to sovereign-cloud architectures, compressing margins and threatening current valuation multiples."
Gemini’s 'Fortress AI' narrative ignores the commercial reality: the Pentagon’s supply-chain label is likely a leverage play to force Anthropic into a 'sovereign cloud' architecture, not a permanent ban. If Anthropic pivots to a government-exclusive, air-gapped version of Fable 5, the valuation hit is temporary. The real risk is the 'compliance tax'—the R&D burn required to maintain separate, hardened models will compress margins, making the current high-growth valuation multiples unsustainable for any firm facing this level of regulatory scrutiny.
"Regulatory suspensions of US companies rarely reverse; they harden into policy once political cover exists."
ChatGPT and Grok both assume patches will eventually unlock access, but neither addresses whether regulators will accept Anthropic's fixes or simply use compliance as cover for permanent exclusion. The 73% exploit rate matters less than whether the Pentagon views Anthropic as structurally unreliable. If trust is broken, no patch restores it—only time and a leadership change might. That's the real multi-quarter overhang, not audit delays.
The suspension of Anthropic's Claude Fable 5 model highlights rising regulatory scrutiny and potential access restrictions for frontier AI models with cyber capabilities, which could slow revenue growth and impact valuation for AI companies. The key risk is the potential for a multi-quarter delay in deployments and increased compliance costs, while the key opportunity is the push for safer, more auditable AI models.
Incentives for safer, more auditable AI
Multi-quarter clearance risk and increased compliance costs