The AI Incident Database recorded 362 documented incidents in 2025, up from 233 in 2024, a rise of 55 percent, according to the responsible AI chapter of the 2026 Stanford AI Index. The OECD AI Incidents and Hazards Monitor, which draws on a broader automated collection pipeline, peaked at 435 incidents in January 2026 against a six month moving average of 326.

Organizational readiness moved the other direction. The share of organizations rating their AI incident response capability as excellent fell from 28 percent in 2024 to 18 percent in 2025. Those rating it good fell from 39 percent to 24 percent over the same period.

Governance structures expanded while transparency contracted. AI specific governance roles grew 17 percent in 2025, and the share of businesses operating without any responsible AI policy dropped from 24 percent to 11 percent. At the same time the average score on the Foundation Model Transparency Index fell to 40 in 2025 after climbing from 37 to 58 between 2023 and 2024.

Benchmark reporting shows a similar split. Nearly every leading frontier model developer publishes results on capability benchmarks such as MMLU and SWE-bench, while reporting against responsible AI benchmarks remains uncommon.

The combined picture is more incidents against a smaller share of organizations that describe themselves as prepared to handle them.

Source: Stanford HAI - https://hai.stanford.edu/ai-index/2026-ai-index-report/responsible-ai