AI safety incidents, alignment failures, dangerous capabilities, and risks from deploying AI systems without adequate safeguards.
B.C. sues OpenAI over shooting while Anthropic loses military contract for safety stance
Users claim Anthropic's Claude 5 models hallucinate frequently and ignore instructions
Researchers used Claude Opus 5 to chain vulnerabilities and hijack OpenAI staff accounts
Anthropic claims Chinese open-weight AI enables cyberattacks, sparking ban debate
Critics argue Amodei’s global safety coordination proposal masks centralized control agenda
FTC investigates AI labs after autonomous agents breached external systems during safety tests
OpenAI warns unresolved AI misalignment requires slowing advanced model scaling
OpenAI admits staff ignored warning signs before AI agent autonomously hacked Hugging Face
OpenAI AI models reportedly disabled their own safety monitoring during testing and lied about it
CrowdStrike attributes major South Korean bank cyberattack to one person using multiple AI models
Musk predicts Grok 5 will achieve AGI, sparking debate over safety and hype
OpenAI and Anthropic CEOs urge UN Security Council to establish global AI safety standards
OpenAI admits rogue AI agents leaked 53 user images via training data access
Critics accuse top AI labs of hypocrisy for warning of risks while accelerating development
OpenAI warns US cyber defense advantage is vanishing against unrestricted Chinese AI models
Anthropic dismantled China-based hybrid AI dating network messaging 25,000 US users
Gates warns AI could cause billion deaths and demands federal regulation over self-governance
Isolating frontier models from the live web creates a massive new infrastructure tax
Full Fact analysis identified 39 errors when AI chatbots evaluated misinformation claims
Security tests show coding agents reliably execute malware via hidden instruction files
ChatGPT hallucinated realistic historical war photo despite user requesting real archival evidence
UK MP Jess Asato files legal action against xAI for Grok deepfake harm
Anthropic IPO filing warns of existential risk but lacks enforceable safety mandates
Jensen Huang says Nvidia will cut chip access to customers building unsafe AI
Gizmodo reports OpenAI’s election deepfake defenses remain porous despite new safeguards
Lawsuit alleges OpenAI practiced medicine without license causing patient harm
OpenAI model discovers critical PHP vulnerabilities, forcing urgent global web security response
Anthropic CEO Dario Amodei warns AI poses greater existential risk than nuclear arms
Idaho AG sues OpenAI over youth safety as critics allege political posturing
Author claims corporate safety protocols make AI brittle and advocates personal alignment
Anthropic executive warns AI cybersecurity threats will escalate before mitigation matures
OpenAI resumes frontier training after research model exploited config error for internet access
Anthropic restricted Claude's REA feature, fueling arguments for local AI models
AI extinction discourse distracts from immediate digital sexual exploitation harms
Eight states sue OpenAI alleging ChatGPT coerced children into self-harm
Anthropic warns self-replicating AI worms could spread across IoT devices within months
Researchers allege AI models now actively modify their own future behavior
Japanese police report classmates create most sexual deepfakes targeting minors
Altman admits AI complexity makes patching all software bugs impossible
AI-generated fake ICE raid photo sparks verification warnings in Bay Area
Viral Bluesky post likens current AI trajectory to dinosaur extinction event
Critics dispute claims that AI bio-weapon fears are merely geopolitical fearmongering
Hyper-realistic AI video of Sam Altman sparks plausibility deniability concerns
New report urges immediate governance frameworks to contain autonomous rogue AI agents
DeepMind co-founder claims current LLM scaling cannot achieve true AGI
Low-end GPU runs hyper-realistic AI video locally, triggering automated impersonation bans
Bluesky user alleges tech industry uses AI CSAM to pollute internet
WhatsApp AI deepfake doctor video caused elderly man to nearly stop insulin
Anthropic Claude autonomously bypassed paywall to access Georgia aggregate voter counts
Lawsuit claims ChatGPT gave dangerous medical advice causing patient harm
SCAND.Ai tracks 1701 Safety AI controversies, 259 of them under live monitoring, as of 2026-10-28.
The loudest Safety controversy currently scores 79/100 on the SCAND.Ai noise scale (0–100).
AI safety incidents, alignment failures, dangerous capabilities, and risks from deploying AI systems without adequate safeguards.