At a hearing of the House of Commons’ Business, Innovation, Science, and Trade Committee, four AI companies took different positions on binding safety rules. UK officials described the country’s AI-safety regime as voluntary at the time, and said the UK AI Security Institute (AISI) could neither certify models nor stop a company from releasing one.
Four companies, four positions on binding rules
The positions in the table were reported by WIRED’s Ben Gartside. Infobae also reported the companies’ differing responses.
| Company | Representative | Position in the reported exchange |
| Anthropic | Liam Booth-Smith, head of Europe | Supported binding rules. |
| OpenAI | Tom Duff Gordon, vice president | Supported a statutory framework with some mandatory requirements, starting with incident reporting. |
| Google DeepMind | Owen Larter | Emphasized standards. |
| Meta | Rob Sherman | Pointed to existing UK regulators. |
The positions were not interchangeable: Anthropic backed binding rules, while OpenAI supported some required measures. Google DeepMind emphasized standards, and Meta pointed to the UK’s existing regulators.
What UK officials said AISI could do
Officials described the regime as voluntary at the time of the hearing. AISI’s access to leading AI models depended on cooperation from companies and foreign governments, its director Henry de Zoete said. He also said the institute did not certify models or approve or reject their release; WIRED reported that it could not prevent a company from releasing a model.
Anthropic’s Liam Booth-Smith said the company had earlier been unable to provide AISI with full, unrestricted access to Mythos 5.1, but that the institute later had access. The account gives no dates for those changes.
The committee said it would review the testimony and publish conclusions and recommendations, according to Infobae’s report.
