2026-W05 · -0.8
During the assessment period, there was no evidence of AI systems acting autonomously without human control or causing severe real-world harm. However, several papers demonstrated techniques to exploit AI systems, such as jailbreaking and adversarial prompting, highlighting ongoing security concerns. There were also documented incidents involving AI, such as a dataset containing child sexual abuse images and adversarial attacks on watermarking defenses. Efforts to improve AI safety frameworks were reported, but governance remains a challenge.
A conventional composite index, not a scientific probability of extinction.