Press

Warning Shots tracks AI’s dangers and breakthroughs as they land on real lives, checked three times a day. Every significant AI incident since 2023 is scored 0–100 by a published formula and linked to its sources. Free to cite, quote and embed.

Key figures

160AI incidents since 2023, each scored and sourced

3 → 13severe incidents, previous 12 months vs. last 12

42%of this year’s real-world cases where the AI itself did the acting

WS-4today’s level: Severe

Figures as of Oct 26, 2026; they change as the index is refreshed. Check the Record for the current numbers before publishing.

How to cite

Short credit, for a chart or a caption:

Source: Warning Shots (warningshots.ai)

Full citation:

Warning Shots. "The AI Incident Index." warningshots.ai, accessed October 26, 2026. https://warningshots.ai/index/

The dataset:

Warning Shots (2026). AI Incident Index dataset [Data set]. https://warningshots.ai/data/incidents.csv

Linking to a single incident? Every incident has its own page, for example the full list. Link to that page so readers land on the sources.

How it works

Each incident is rated on five factors (harm, reach, autonomy, novelty and irreversibility) and combined into a 0–100 score by a formula published on the site. Claims about live events are graded by proof: Tried (attempts logged), Proven (a test system broken into, no real victim yet), Hit (a real victim confirmed). The three are never blurred. Forecasts are dated and never edited after they resolve, and every change is logged.

Screenshots

The front page: the day’s level, the running story and the headlines

The Record: every incident since 2023, scored and sourced

Logo and badge

Logo mark, PNG 800 px Logo mark, SVG Today’s share card, PNG Live level badge for your site

Data

The full index is downloadable: CSV · JSON. New reports arrive by RSS.

Contact

Open an issue on GitHub; corrections are answered first and logged in the public changelog.