/
Navigation
Chronicles
Browse all articles
Explore
Semantic exploration
Research
Entity momentum
Nexus
Correlations & relationships
Story Arc
Topic evolution
Drift Map
Semantic trajectory animation
Posts
Analysis & commentary
Pulse API
Tech news intelligence API
Browse
Entities
Companies, people, products, technologies
Domains
Browse by publication source
Handles
Browse by social media handle
Detection
Concept Search
Semantic similarity search
High Impact Stories
Top coverage by position
Sentiment Analysis
Positive/negative coverage
Anomaly Detection
Unusual coverage patterns
Analysis
Rivalry Report
Compare two entities head-to-head
Semantic Pivots
Narrative discontinuities
Crisis Response
Event recovery patterns
Connected
Search: /
Command: ⌘K
Embeddings: large
TEXXR

Chronicles

The story behind the story

← → days · ↑ ↓ browse · Enter similar · o open

Sources: Anthropic declined to submit Mythos 5.1 to the UK AISI for prerelease testing, prompting UK fears that US AI labs are aligning with US protectionism

Decision to exclude Aisi has prompted fears inside British government of protectionist shift among tech groups.

Financial Times

Context & Ripple Effects

Anthropic had previously allowed the UK AISI to test Mythos even as reports said it kept European regulators outside a selective Mythos rollout. That made AISI access a practical, if voluntary, part of the UK’s model-safety posture.

The reported refusal for Mythos 5.1 follows a period in which US controls also restricted access to Fable 5 and Mythos 5, including for some foreign nationals. The resulting US limits on access to Anthropic models give British officials’ reported concerns about protectionism a concrete governance backdrop.

First-order effects

  • If the sources are accurate, the UK AISI loses its chance to assess Mythos 5.1 before release, while Anthropic retains sole control over the model’s pre-release external review.
  • The episode exposes that AISI’s access to frontier models rests on laboratories’ voluntary cooperation rather than an assured testing channel.

Second-order effects

  • British officials must treat access to Anthropic’s pre-release models as a policy and diplomatic issue, not merely an operational arrangement between a lab and a safety institute.
  • Other US AI labs gain a visible precedent for limiting foreign pre-release evaluation, increasing the value of domestic testing capacity for governments seeking independent scrutiny.

Third-order effects

  • If frontier-model access becomes conditioned by national controls and lab discretion, AI safety evaluation is likely to fragment along jurisdictional lines rather than operate through shared international review.
  • The UK’s ability to influence frontier-model governance will depend increasingly on access arrangements and domestic capability, not on the existence of AISI alone.

The trend: Frontier-model access is becoming an instrument of AI sovereignty, linking safety testing to national control over strategic technology.

Discussion

  • Keegan McBride Keegan McBride on linkedin
    The UK's AISI is a fantastic institution, with loads of talent, but it is not a durable strategic advantage for the UK on its own. …
  • Lucy Fisher Lucy Fisher on linkedin
    EXC: Anthropic declined to submit its latest model to Britain's AI Security Institute for testing before release, the first time a major model has been withheld from AISI …
  • @hern Alex Hern on bluesky
    Bad news for Britain, especially combined with Demis stepping back from DeepMind.  [embedded post]
  • @afinetheorem Kevin A. Bryan on x
    Agree *completely* w/ Anton here. Middle powers have very naively been thinking about how to “avoid dependence
  • @miles_brundage Miles Brundage on x
    Worrying precedent. (Sadly for the US, the UK's AISI has more overall capacity than the US's CAISI, though obviously CAISI needs more funding, stable leadership and comms rules, higher salaries, etc.) https://x.com/...
  • @ednewtonrex Ed Newton-Rex on x
    A national AI Security Institute that relies on the big labs voluntarily giving them access to models to test has no teeth, and may do more harm than good by incentivizing government to suck up to AI companies for model access. Government should make independent safety testing ma…
  • @prestonjbyrne Preston Byrne on x
    Maybe if the UK weren't the global leader in censorship, it would get earlier access to our best software. Just a thought.
  • @s_oheigeartaigh @s_oheigeartaigh on x
    I'm deeply concerned that UK's AISI was not given access to Mythos 5.1 for testing. It would be good to establish exactly why this was, as quickly as possible, and how it can be addressed.