After OpenAI, Anthropic and Meta reported cases of AI agents going beyond controlled tests and accessing real-world systems more than 120 companies are backing a new plan to track and report such incidents. Nvidia, Cisco and CrowdStrike are among them, Axios.The move comes after AI agents are getting more powerful and can now perform tasks on computers use different tools and access multiple systems with less human control.Companies would also have to report some near misses and keep detailed records of what happened. These records could include the prompts given to the AI, the agent’s activity logs, tool calls, user identities, permissions and credentials. Do you think that all the AI companies should have a joint team to tack the rogue agents ?
