AI Scientific Panel: key recommendations on frontier AI safety
After the special meeting of October 9, a Commission page sets out key recommendations from the AI Scientific Panel on the safety of the most advanced models.
Source of truthdigital-strategy.ec.europa.eu/…/special…We have covered the special meeting of October 9, held with Commission Executive Vice-President Henna Virkkunen. A Commission page dated October 12 now sets out key recommendations the panel’s chair and vice-chair presented on frontier AI safety and security. In particular, they recommend steps to:
- govern the risk from internal models while they are being developed and tested;
- strengthen EU capabilities to detect and investigate loss-of-control incidents, for example through a task force;
- set up an international initiative for information-sharing, common requirements and joint crisis response;
- apply closer scrutiny to training techniques that make models harder to monitor.
These are recommendations, not decisions. The Commission will send requests for information to the companies behind the models involved. The page does not name the companies.
Related stories
Microsoft: Storm-0940 group automated attacks with AI agents
According to Microsoft, Storm-0940, which it assesses as China-state aligned, used the open-source ARTEX framework to automate attacks targeting hundreds of organizations; most of the targeting was not successful.
LiteLLM: malicious version 1.107.0 published to PyPI and Docker
According to LiteLLM, version 1.107.0, pushed on October 21, contained malware that steals credentials. LiteLLM advises anyone who installed it on a host with outbound internet access to assume the data was exfiltrated and to rotate keys and secrets.