27JUL

Anthropic AI ban enters its second week

3 min read

10:02UTC

Anthropic's Fable 5 and Mythos 5 have been switched off for foreign users for eight days. The promise of restoration 'within days' has not been kept.

← Back to SAP freezes R&D headcount as others deny AI Jump to analysis ↓

EconomicDeveloping

Key takeaway

Washington switched off a major AI firm's models on security grounds while its identically-flawed rival stayed on sale.

Anthropic's Fable 5 and Mythos 5 models have NOW been suspended for roughly eight days, after Commerce Secretary Howard Lutnick ordered the company to bar foreign nationals from access on 12 June . Anthropic is a US AI developer; the two models power its consumer and frontier products. The promise that service would return "within days" has not been kept. Senior Anthropic staff met Department of Commerce officials in Washington this week to negotiate, House members publicly demanded answers on 18 June, and President Trump, at the G7 summit of advanced economies, said the talks were "going fine" ¹.

Lutnick's letter cited the risk of military-intelligence use by China and Russia. Axios reported a different sequence: relations ruptured after Anthropic refused to let the US military use its models for domestic surveillance and fully autonomous weapons, and the government then placed the company on a national-security blacklist ². Anthropic counters that OpenAI's GPT-5.5 carries the identical jailbreak vulnerability and remains on public sale. A security action that hits one lab and spares its rival, after that lab funded a pro-regulation campaign , reads to critics as political retaliation rather than threat mitigation. Lutnick's office maintains the same export-control rule would apply to any vendor carrying the vulnerability.

The rupture went transatlantic. The G7 summit at Evian, held 15 to 17 June, failed to reverse the controls ³. European governments arrived unsettled that a model they had deployed could be turned off worldwide by one letter, with no appeal and no notice to customers. Brussels has folded the episode into its European Technological Sovereignty Package, presented on 3 June, which bundles a chips programme with a Cloud and AI Development Act meant to wean the bloc off US infrastructure. A UK exemption was sought at Evian and refused.

Deep Analysis

In plain English

Anthropic makes two powerful AI models called Fable 5 and Mythos 5. On 12 June, the US government ordered Anthropic to stop letting people outside the US use them. The suspension has now lasted more than a week. Axios reported that the real reason is a falling-out: the US government wanted to use Anthropic's AI for domestic surveillance and to control autonomous weapons, and Anthropic refused. The government then put Anthropic on a national-security blacklist. Anthropic has pointed out that a rival product, OpenAI's GPT-5.5, has the same security flaw the government cited; and GPT-5.5 is still available. The G7 leaders met in France and could not agree on a way to reverse the US controls.

Deep Analysis

Root Causes

The Anthropic suspension reflects two underlying structural tensions that predate this specific action.

First, the 2018 Export Administration Regulations (EAR) give the Commerce Department broad authority to restrict technology transfers to foreign nationals without defining AI-specific thresholds. Commerce's application of EAR to a software-as-a-service model; where the 'transfer' is API access, not physical export; is legally novel and untested in federal court. The government is using a regulatory framework designed for hardware and encryption to control AI model access.

Second, the Anthropic case makes visible a structural conflict between the national-security state's demand for exclusive control over frontier AI capabilities and the commercial model of frontier AI firms, which depends on broad access (including international researchers and enterprise customers) to generate the revenue that funds further model development. These two interests cannot both be satisfied simultaneously.

What could happen next?

Precedent
A sustained suspension of a major US frontier AI model on national-security grounds would establish that the US executive can effectively nationalise AI access as a geopolitical tool, prompting European governments to accelerate sovereign AI model development.
Medium term · Assessed
Risk
If courts reject Commerce's EAR application to SaaS API access, the legal basis for AI export controls collapses, leaving the government with no standing mechanism to enforce similar restrictions on future refusals.
Short term · Reported
Consequence
Anthropic's identification of GPT-5.5's identical jailbreak as a public counter-argument establishes an asymmetry precedent: competitors who maintain government relationships are not subject to the same access restrictions regardless of their security posture.
Immediate · Assessed

Source Landscape

This story draws on centre-left-leaning sources from United States and France

United States×2

France

LeftRight

Primary parallel: IBM's 1969-1982 antitrust entanglement with the US Department of Justice offers a structural parallel: a market-dominant technology firm that refused specific government use cases found its commercial relationships complicated by regulatory action that was officially framed on different grounds.

The IBM case lasted 13 years before the government dropped the suit. The Anthropic action appears faster and more targeted, but the pattern; government pressure through non-antitrust regulatory levers after a commercial disagreement; is structurally similar.

Counter-parallel: The 2020 TikTok executive order, which attempted to ban the app on national security grounds, provides a counter-case where the legal basis proved fragile: courts repeatedly blocked enforcement, and the action was eventually resolved through negotiation rather than ban. If Anthropic pursues litigation, the TikTok precedent suggests executive export-control actions face a higher judicial bar when the national security rationale is disputed.

Consensus view: The Washington Post's reporting on the 18 June House demands establishes bipartisan legislative anxiety about the export control mechanism. The core concern, as documented by Euronews at the G7 summit, is less about Anthropic specifically and more about whether the US executive can effectively impose a 'kill switch' on frontier AI access globally; a precedent that European governments regard as an infrastructure sovereignty threat.

Counter-view: Axios reported that the rupture originated in Anthropic's refusal to permit US military use for domestic surveillance and autonomous weapons, framing the action as Anthropic's choice, not a government overreach. From this reading, the Commerce Department's national-security blacklist is a normal state response to a commercial entity declining a procurement condition; not categorically different from a defence contractor declining a classified contract.

Key tension: Whether Anthropic's parallel argument; that GPT-5.5 carries the identical jailbreak vulnerability and remains on sale; constitutes evidence that the export control is punitive rather than security-motivated, and whether Congress will pursue that thread when demanding answers.

First Reported In

Update #14 · The AI layoffs nobody is counting

Washington Post· 20 Jun 2026

Read original →

Causes and effects

Caused by

Washington pulls a live AI model

The Lutnick directive of 12 June caused the ongoing suspension that entered its second week, triggering House demands, G7 discussions, and Washington negotiations.

Occurred 12 Jun 2026

Read story →

OpenAI heads to market losing money

OpenAI's IPO filing for a September listing is referenced as context for why the selective enforcement of the Anthropic ban reads as political retaliation.

Occurred 9 Jun 2026

Read story →

This Event

Anthropic AI ban enters its second week

A deployed US AI model can now be withdrawn worldwide by a single regulatory letter, with no appeal and no customer notice.

Different Perspectives

European Commission

The European Commission's draft Annex III guidelines, closed for comment on 23 July, treat algorithmic scoring in recruitment, pay and termination as high-risk regardless of whether a human signs off, echoing Spain's Audiencia Nacional ruling 101/2026 on concealed scheduling algorithms. Brussels is shifting the fight from counting AI job losses to assigning legal liability for the tools themselves.

Office for National Statistics

The Office for National Statistics recorded UK vacancies rising to 712,000 on 21 July, the first quarterly increase this beat has tracked, with payrolled employment down 85,000 on the year against May's 210,000 fall. The bulletin names no AI cause anywhere, and that is the point: nothing in the release confirms the displacement story it gets cited to support.

Christian Klein, SAP

Christian Klein told investors on 23 July that SAP's research headcount will not grow for twelve months because AI agents and their token costs are absorbing the work, not because SAP is cutting jobs. He frames it as commercial arithmetic: the cost of AI-assisted coding tokens plus the salaries specialist AI hires command, not people being replaced by machines.

Betsey Stevenson, University of Michigan

Betsey Stevenson argued that the 187,000 jobless-claims reading describes a market that hires little and fires little, not one AI is emptying. She said the real damage hides in eligibility rules and suppressed job postings, not in the headline layoff counts employers keep denying.

Comisiones Obreras, UGT and Concentrix's A Coruña works committee

Comisiones Obreras, UGT and Concentrix's A Coruña works committee blamed Microsoft's push toward AI self-service for the 80 redundancies unions signed off on 22 July, not unavoidable business cause. A second Coruña procedure covering 80 more jobs runs to a 31 August deadline, and the unions want the state, not the employer, setting the pace of AI-driven cuts.

Stanford's 'We Must Act Now' signatories

More than 200 academics, including 16 Nobel laureates, published a 13 July letter warning of AI-driven labour disruption, citing Daron Acemoglu's NBER estimate that AI's total factor productivity gain stays under 0.66% over ten years. The letter's own cited economics sit well below Goldman Sachs Research's 1.5-percentage-point estimate published the same week.