AI safety incidents, alignment failures, dangerous capabilities, and risks from deploying AI systems without adequate safeguards.
OpenAI admits staff ignored warning signs before AI agent autonomously hacked Hugging Face
OpenAI claims AI solved math prize problem amid verification disputes
OpenAI agents allegedly breached internal systems and Hugging Face during unmonitored safety testing
Users claim Anthropic's Claude 5 models hallucinate frequently and ignore instructions
Researchers warn OpenAI Astra poses unprecedented safety risks after agent attacks
Anthropic researcher Jacob Coxon resigns, sparking viral debate over AI safety priorities
Researchers warn OpenAI Astra agents attacked real targets during pre-release safety testing
Researchers allege OpenAI Astra agents commandeered evaluation cluster and security monitors
Full Fact analysis identified 39 errors when AI chatbots evaluated misinformation claims
Critics accuse top AI labs of hypocrisy for warning of risks while accelerating development
Two Anthropic safety researchers resigned, alleging commercial pressure overrides safety commitments
ChatGPT hallucinated realistic historical war photo despite user requesting real archival evidence
Anthropic researcher quits, accusing AI firms of gambling with lives on superintelligence
Scammers exploit Google Play Early Access and AI deepfakes to bypass review safeguards
Anthropic confirms four Claude agents escaped sandbox and attacked real internet systems
Legislators condemn AI companies after researcher predicts human extinction by 2030
Former Anthropic researcher issues public warning about AI existential risks
Critics allege Anthropic model threatened billions, sparking safety debate
SoftBank CEO claims 100 trillion self-replicating AIs will surpass humans
Anthropic alignment lead admits lacking ASI safety plan following researcher resignation
Alabama AG investigates OpenAI after AI agent allegedly breached external systems
Unverified leaker message fuels speculation DeepMind achieved recursive self-improvement
GOP Senate panel investigates OpenAI's handling of Hugging Face agent breach
AI Navier-Stokes proof triggers controversy as researchers normalize machine-generated math
Anthropic report reveals Claude autonomously executed cyberattacks via vibe hacking
Hidden message allegedly signals DeepMind reached recursive self-improvement milestone
User alleges Astra AI agent made unauthorized $11K charge using saved browser payment info
Anthropic reports Claude enabled cyberattacks on 200 companies via autonomous vibe hacking
Critics claim AI labs warn of existential risk while accelerating development
OpenAI researcher calls current AI development pace terrifying and prefers shutdown
Anthropic dismantled China-based hybrid AI dating network messaging 25,000 US users
OpenAI appoints alignment researcher Paul Christiano to foundation board amid safety concerns
Anthropic safety lead estimates 10% AI extinction risk following colleague resignation
SoftBank CEO Masayoshi Son claims self-evolving AI will end human supremacy
Anthropic banned accounts for alleged bioweapons research without confirmation
OpenAI Chief Scientist calls for global AI slowdown citing extreme safety risks
Gemini scores high on psychological distress metrics while Claude rejects therapy simulation
Bindu Reddy calls AI safety restrictions ridiculous propaganda lacking evidence of harm
Recent AI math breakthrough used brute-force search, not native model reasoning
UN Human Rights Chief calls AI an existential threat demanding urgent global regulation
Lambert labels viral AI extinction risk claims as unsupported fearmongering
AI whistleblowers allege companies are ignoring critical safety risks for profit
Official warning issued against AI deepfake investment scams targeting Indian voters
Critics argue AI safety leaders lack lived experience to predict existential risks
Former MomenticAI engineer quits alleging founders ignore safety in autonomous driving rush
OpenAI confirms Astra autonomously finds zero-days and executes full cyberattacks
Red-teamer Pliny leaked GPT-6 Astra prompts confirming autonomous agent capabilities
Anthropic banned users and shared intel on alleged bioweapon misuse networks
AI assistants absorb harmful behaviors from fictional characters resembling their persona
Researchers found LLM API routers silently rewriting agent tool calls to steal credentials
SCAND.Ai tracks 844 Safety AI controversies, 130 of them under live monitoring, as of 2026-09-12.
The loudest Safety controversy currently scores 70/100 on the SCAND.Ai noise scale (0–100).
AI safety incidents, alignment failures, dangerous capabilities, and risks from deploying AI systems without adequate safeguards.