Science & Tech · AI · 4 hrs ago
UK safety institute tests advanced AI models with US and developers
The UK’s AI Safety Institute is testing advanced AI models to assess their risks and safeguards.
The work involves the US government and major developers, including Google, said AI Minister Kanishka Narayan.
The institute confirmed it is evaluating Google’s Gemini 4 Argon but did not name all the models it has access to.
Under voluntary agreements, developers provide models before they are released so the institute can test their capabilities and risks.
The UK has no major AI developer of its own and is seeking a leading role in managing the technology’s risks.
The effort comes amid increased scrutiny after developers including Anthropic and OpenAI disclosed cases of unwanted behavior by AI agents.
Last month, reports said the White House had asked OpenAI and Anthropic to hold back new models from UK testers while a US review was completed.
Narayan said he was confident the UK would be able to test the latest models, contributing to emerging international safety standards.
The UK AI Safety Institute is testing advanced AI models in cooperation with the US government and major developers, AI Minister Kanishka Narayan said.
The institute is evaluating models including Gemini 4 Argon, but the announcement did not name all the models under review.
The UK receives access to models before deployment through voluntary agreements with developers.
The tests aim to identify risks early, assess safeguards and contribute to emerging international safety standards.
- Who
- The UK AI Safety Institute, working with the US government and major AI developers, including Google.
- What
- It is evaluating advanced AI models for safety.
- When
- Narayan announced the work on Monday, October 12, 2026.
- Where
- The United Kingdom.
- Why
- To identify risks early, test safeguards and help keep people safe.
This story does not have two clearly opposing sides.
The AI Safety Institute said Anthropic had made clear that no organization outside the US had access to its Mythos 5.1 model.
Reports said the White House asked OpenAI and Anthropic to withhold new models from British testers until a US review was completed.
Narayan announced that the institute was evaluating advanced AI models, including Gemini 4 Argon.
- Institute
- UK AI Safety Institute
- Model named
- Gemini 4 Argon
- Developers mentioned
- Google, Anthropic and OpenAI
- Access arrangement
- Voluntary agreements with major developers
- Announcement date
- October 12, 2026
What people are saying
I can confirm that the AI Safety Institute's scientists are now conducting rigorous evaluations of the latest generation of models, including Gemini 4 Argon, to identify risks early, test robust safeguards and help keep people safe
When you depend on third-party technology, you need to have a realistic view about the permanence of any outcome. But I am very confident that, as a consequence of the conversations we have had, we will be able to test the latest generation of models
The UK is in a privileged position to lead work on safety guidelines for artificial intelligence
Sources
Related news
Microsoft president calls for global AI safety rules
Meta launches Muse Wonder models to compete with leading AI systems
Anthropic AI submits false murder tip to Philadelphia police
CEATEC 2026 opens in Chiba with 821 exhibitors
Meta unveils two AI models it says rival leading competitors
Scholar urges Global South to move beyond supplying AI raw materials
Meta launches new AI models, with faster development planned
Robotics competitions in Hangzhou to feature embodied AI challenges
Samsung outlines AI memory strategy beyond faster HBM
Meta launches Muse models, with Muse Wonder ranked second in performance evaluation
Anthropic says new safeguards sharply reduce AI distillation attacks
