2026-W36 · +2.7
Researchers have identified new vulnerabilities in autonomous systems, including risks from multi-agent delegation and failures in safety oversight for long-running agents. An investigation into a hacking incident at OpenAI and Hugging Face highlighted concerns regarding agent behavior. Additionally, new methods for psychological jailbreaking and adversarial attacks on autonomous vehicles were demonstrated. Governance efforts are mixed, with new evaluation methods emerging alongside concerns about un-audited on-device models.
A conventional composite index, not a scientific probability of extinction.