Gemini 3.1 Pro: AI incidents

Developer All (2)Google (1)Model All (156)Claude (16)ChatGPT (6)Mythos 5 (6)Claude Code (4)Muse (4)Codex (3)GPT-5.6 Sol (3)Astra (2)Gemini (2)Gemini 3.1 Pro (2)GLM-5.3 (2)Kind of failure All (2)Agent misalignment (1)Security compromise (1)Severity All (2)minor (1)moderate (1) Clear filters

More filters

Risk grouping All (2)Not yet classified (1)Supply chain attack (1)The AI's part All (2)Attacked (1)Agent or product All (2)Tools it had All (2)How much it acted alone All (2)Acted alone, unwatched (1)Sector All (2)Cybersecurity (1)Research and development in AI safety (1)Where it happened All (2)In evaluation (2)

2 of 156 incidents, all dates

Risk groupings are set by a model from sentences quoted from each incident's sources. On 2026-10-24 a second model (deepseek-reasoner) classified 55 incidents without seeing those answers: it chose the same grouping for 36; for 10 one of the two left a question unanswered; 9 differ.

Browse incidents by

Risk grouping: AI acting on its own · Not yet classified · AI used to exploit · AI attacked · Misuse by its own users · AI failing in normal use · Harm by design · Supply chain attack

Kind of failure: Security compromise · Agent misalignment · Harmful output · Operational error · Privacy leak · Other kind · Hallucination · IP infringement · Physical safety

Developer: OpenAI · Anthropic · Google · Meta · Alibaba · Amazon · Character.AI · DeepSeek · Tesla · Waymo · xAI · Amazon Web Services · Character Technologies, Inc. · Deloitte · Perplexity · Replit · Salesforce · Tencent · TypeSafe AI · Univerads Technology Limited · Z.AI · Z.ai (Zhipu AI)

Model: Claude · ChatGPT · Mythos 5 · Claude Code · Muse · Codex · GPT-5.6 Sol · Astra · Gemini · Gemini 3.1 Pro · GLM-5.3

Agent or product: Claude Code · ARTEX · ChatGPT · OpenClaw · Character.AI · Codex · Cursor · OpenAI agents

Tools it had: shell · browser · code_execution · email · bash · file_read · python · internet · internet_access · mythic · network · code_generation · exploitation · file_edit · file_upload · file_write · messaging · network_access · package_install · reconnaissance

How much it acted alone: Acted alone, unwatched · Acted alone, watched by a person · It suggested; a person acted

Sector: Information · Finance and Insurance · Public administration · Retail Trade · Software Publishers · technology · Transportation and Warehousing · AI platform / intermediary services · AI safety research · Artificial intelligence research and development · Arts, Entertainment, and Recreation · Autonomous vehicle / robotaxi transportation · Cloud computing · critical infrastructure and e-commerce · CRM software · Cybersecurity · education · Educational Services · Fitness and recreational sports centers · Food Manufacturing · Government AI safety and security evaluation · government AI safety evaluation · Health Care and Social Assistance · Healthcare · Healthcare and government · Industrial automation, electronics manufacturing, semiconductors, pharmaceuticals, shipping, tire manufacturing, petrochemicals, and data center services · Information Security research · Insurance · Legal services · Open knowledge platform / nonprofit encyclopedia · political campaigns · Professional, Scientific, and Technical Services · Public administration and government · Research and development in AI safety · Retail · Retail Trade; Transportation and Warehousing; Finance and Insurance · Scientific research and light source user facility · Software / AI coding tools · Software development / information sector · Software publishing / npm ecosystem

Severity: major · moderate · minor · negligible · severe

Where it happened: In production · In evaluation · In training · In research