US Allowed AI Developers to Share Models With Britain
The Trump administration reversed a ban on sharing advanced AI models with Britain’s AI Security Institute for safety testing.
Updated on Oct. 12, 2026 in Artificial Intelligence
Live Poll
Should the US government allow international organizations to conduct safety testing on American AI models?
The United States government has authorized American AI developers, including OpenAI and Anthropic, to share new artificial intelligence models with the British AI Security Institute. This policy reversal allows British researchers to conduct safety testing on emerging systems prior to their public release.
Why it matters
The sharing of proprietary models with international partners enables cross-border collaboration on safety testing for advanced artificial intelligence. It marks a shift in how U.S. developers handle internal security vetting in coordination with international safety institutions.
The British AI Security Institute—a body formed following the 2023 AI safety summit—will now receive access to models that were previously restricted. This allows for standardized safety evaluations by a foreign institution, building on the security requirements of the U.S. AI task force.
The players
OpenAI
A developer of large language models and proprietary artificial intelligence systems.
Anthropic
An AI research and development company focused on building steerable and safe AI models.
British AI Security Institute
A British government entity tasked with evaluating the safety and risks of advanced artificial intelligence.
The details
The British AI Security Institute conducts rigorous safety testing to identify vulnerabilities in large-scale AI architectures before they reach the public. By allowing American firms like OpenAI and Anthropic to share these models, the U.S. government is integrating international oversight into its domestic safety workflow. This process operates alongside the U.S. AI task force, which was launched on October 9, 2026, to monitor and document model security incidents.
Timeline
2023: Britain established the AI Security Institute.
October 9, 2026: The U.S. AI task force was launched.
October 11, 2026: The policy reversal was reported.
The Tech Race
The U.S. decision follows the collaborative precedent set by the 2023 AI safety summit. It formalizes an international model-sharing pipeline, expanding the reach of safety evaluations beyond domestic institutions.
This development standardizes safety testing processes, which may influence the release timelines of new AI tools used by the public. Users will likely see more rigorous safety disclosures as international vetting becomes a standard prerequisite for high-stakes model deployments.
The takeaway
The normalization of international model sharing highlights an increasing reliance on cross-border cooperation for artificial intelligence safety. Stakeholders should watch for future reports issued by the British AI Security Institute as the first models undergo this multi-national testing process.
Further reading
For broader context on how safety standards are evolving, see Artificial Intelligence.
Live Poll
Should the US government allow international organizations to conduct safety testing on American AI models?
