OpenAI has thrown a curveball into its plans for the release of GPT 5.6, its latest artificial intelligence model, after receiving a request from the US government to slow down the rollout. The company behind ChatGPT will now implement a phased release, starting with a limited preview for a select group of partners that need approval from the US authorities on a case-by-case basis.
This measured approach echoes Anthropic's strategy for its Mythos model earlier this year. Initially introduced with a staggered release programme, Anthropic eventually withdrew the technology entirely due to pressure from the US government, which cited concerns over its potential use in cyber-hacking. The UK's AI security body had hailed Mythos as a significant leap forward, but also highlighted the increasing power and risks associated with these technologies.
Discussions between OpenAI and two key US agencies - the Office of the National Cyber Director and the Office of Science and Technology Policy - have prompted this change in plans. While OpenAI would prefer to adopt its own release schedule, it has agreed to work with the government and industry stakeholders on a more sustainable approach for future AI launches. Reports suggest that Howard Lutnick, a former commerce secretary under Donald Trump, intervened to secure additional approvals from other agencies before any limited release could proceed.
This shift in the US government's stance on AI regulation is significant. Just last month, the president signed an executive order introducing a voluntary framework for federal vetting of powerful new AI models before their release. This move marks a departure from previous warnings from senior White House figures about excessive regulation potentially stifling the transformative AI industry.
For businesses and consumers in the UK, these developments have major implications. While the immediate focus is on US-based releases, the interconnected nature of the tech industry means that global considerations for safety, security, and responsible deployment will likely come into play. The UK's regulatory bodies, such as the ICO, and discussions around frameworks like the EU AI Act, are already adapting to these changes.