iTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
The US reportedly reversed a request that AI developers not give the UK early access to new models for safety testing, according to The Telegraph, as reported by Reuters on 11 October 2026. The report concerns potential pre-release access; it does not establish that any particular model was tested or guarantee access to future models. The UK and US have previously described plans for AI safety cooperation, but their published agreements do not confirm this reported change.
What the reported reversal means
Reuters, citing The Telegraph, reported that the Trump administration had reversed a request for developers such as OpenAI and Anthropic not to share new AI models with the UK AI Security Institute (AISI) before release. Reuters said spokespeople for the administration and the two companies did not immediately respond to requests for comment. The October 2026 report is therefore attributed to The Telegraph through Reuters; the official documents available here do not independently confirm the reversal. Reuters via Investing.com, 11 October 2026.
In practical terms, the reported change concerns whether developers may provide the UK institute with access to unreleased models for safety testing. The report does not specify the legal instrument or operational process behind the change, which US office issued the earlier request, or the precise models and developers covered. Nor does it establish that access is guaranteed for any future release.
What the UK AI Security Institute does
The AISI is a UK government institute established after the 2023 AI Safety Summit. Reuters describes its role as testing advanced AI models to assess their capabilities and risks of misuse. Testing before a model is released can give evaluators an opportunity to examine capabilities and potential risks earlier than testing a publicly accessible model, but the October report does not identify a completed pre-release evaluation. Reuters via Investing.com.
#1 Best Overall
How the report differs from existing UK-US agreements
The 2024 AI safety memorandum
In April 2024, the UK and US announced a memorandum of understanding on AI safety. Their stated plans included developing shared approaches to model evaluations, conducting at least one joint evaluation of a publicly accessible model, collaborating on technical research, exploring personnel exchanges, and sharing information subject to national laws and contracts. The commitment to conduct a joint test was a stated intention—not evidence here that the test was completed. The memorandum addressed cooperation, including evaluation of a public model; it does not establish a guarantee of early access to unreleased models. UK Government: Collaboration on the safety of AI.
The UK government’s announcement of the partnership described plans to align scientific approaches, develop evaluation suites, share capabilities, and conduct at least one joint evaluation of a publicly accessible model. It also framed AI safety as a shared concern, but those 2024 statements were about the partnership at that time—not the reported October 2026 change. UK Government: UK & United States announce partnership on science of AI safety.
Rank #2
The 2025 Technology Prosperity Deal memorandum
A memorandum published in September 2025 described intended cooperation between the US Center for AI Standards and Innovation and the UK AISI. Its AI provisions cover secure AI innovation, model metrology and standards, understanding advanced model capabilities, and exchanging talent. The memorandum explicitly says it does not create legally binding obligations. It therefore describes a framework for intended cooperation, not an enforceable promise that UK evaluators will receive any particular unreleased model. UK Government: Memorandum of Understanding regarding the Technology Prosperity Deal.
Do these 3 things before closing this tab:
1Scan for outdated or missing drivers - takes under a minute2Repair Windows errors before they cause bigger problems3Fix the driver behind crashes, sound loss and screen glitchesWhat is confirmed—and what remains unconfirmed
- Reported: The US administration reversed an earlier request restricting developers’ sharing of new models with the UK AISI before release, according to The Telegraph as reported by Reuters.
- Documented as intended cooperation: The 2024 memorandum describes shared evaluation methods, research, information-sharing subject to law and contracts, and at least one joint test of a publicly accessible model. The 2025 memorandum describes further planned institute cooperation and says it is non-binding.
- Not established by these sources: The formal text or legal basis of the earlier request, the operational details of the reported reversal, a guarantee of access to future models, or confirmation that a specific pre-release test took place.
The UK Parliament also published a written statement on the UK-US partnership in AI safety research in April 2024. It is a record of the partnership, not confirmation of the October 2026 report. UK Parliament: written statement on the UK-US Partnership on AI safety research.
Quick Recap
Best Value
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
