OpenAI and Anthropic Tell British Lawmakers: Regulate Us (www.wired.com)

OpenAI and Anthropic have urged British lawmakers to implement binding AI safety regulations during a parliamentary hearing, asserting that voluntary commitments are insufficient for ensuring safety in AI development. This call for new legislation highlights a significant divide within the AI industry; while OpenAI and Anthropic advocate for a statutory framework, their competitors Google DeepMind and Meta oppose such measures, preferring to focus on existing standards. Anthropic's European head emphasized the need for mandatory rules, while OpenAI's vice president acknowledged the importance of incident reporting as a starting point for regulation. The hearing underscores growing concerns about the potential dangers of AI, with MI5's director warning about the misuse of AI in bioweaponry. The British AI Security Institute disclosed that its current testing regime relies entirely on voluntary cooperation from companies, revealing gaps in regulatory oversight. Former researchers from prominent AI firms expressed skepticism about the industry's commitment to self-regulation, indicating a need for stronger governance. The committee plans to explore effective regulatory frameworks, hoping the G20 presidency will facilitate a broader international convention on AI safety that can operate independently of political fluctuations in the U.S.