Anthropic co-founder Jack Clark called for countries to recognize one another’s AI safety tests. At the October 16 IMF panel in Bangkok, he proposed common standards anchored by national AI security institutes.
The approach described by The Nation would avoid repeating the entire testing process in every country. Clark presented a proposal, not an existing agreement.
Reuters also reported that Clark opposed leaving safety requirements wholly voluntary. He pointed to his company’s regular testing and disclosures.
Clark sharpened Georgieva’s racing analogy by suggesting the cars might be driving at night without brakes. The warning states his position; standards and arrangements for recognizing results remain unspecified.