Britain considers reporting rules for AI safety incidents
英国、AI安全上の事故の報告義務を検討テクノロジー
- メディア
- The Guardian
- 国
- 🇬🇧 United Kingdom
日本語の要約
英語音声
Voice: Sophie
英語本文
Mandatory reporting of AI safety incidents is under consideration in Britain. Kanishka Narayan discussed the issue at Imperial College London on October 12, 2026. He said mandatory reporting was being considered for the Cyber Security and Resilience Bill. Narayan also described Dan Jarvis’s plans for a government cyber defence programme, bringing the legislative discussions alongside the government’s planned security work.
A parliamentary AI safety hearing is scheduled for October 13 before the Business, Innovation, Science and Trade Committee. OpenAI and Anthropic are scheduled to take part. Narayan also says the AI Security Institute uses Gemini and identifies AI safety as a focus for the 2027 G20 presidency. His account connects current security work with the planned international agenda.
Narayan warned about an AI intrusion targeting Australian Medicare in June 2026 that was connected with OpenAI. However, the Australian government has not established that personal medical records were accessed. That distinction matters when describing the incident: a reported intrusion and confirmed access to sensitive records are separate claims, and the latter remains unestablished.