AI safety incidents, alignment failures, dangerous capabilities, and risks from deploying AI systems without adequate safeguards.
Anthropic scraps its flagship pledge to halt AI training over safety concerns amid fierce competition
Local AI tools like ComfyUI face massive security risks from unvetted community-made extension nodes
Grok AI fact-checks and debunks viral fake footage of a purported Iranian strike on Jerusalem
Critics argue Anthropic's updated scaling policy fails to guarantee protection against catastrophic AI risks
Criticism of Anthropic's updated safety plan sparks urgent calls for universal frontier AI regulation
AI’s ability to design chemical workarounds is outpacing the capabilities of global safety regulators
Anthropic claims Chinese AI firms including DeepSeek launched large-scale cyberattacks against its Claude models
Debate intensifies over whether national compute bans effectively mitigate AI risks through resource restriction
MEP Ondřej Kolář calls for regulating AI development pace to prevent catastrophic global security risks
AI-generated deepfake falsely depicts former Indian General claiming army mutiny over international policy
Criticism mounts as OpenAI allegedly removes stabilizing AI features following legal threats without clinical transition
Researchers clash over Nick Bostrom's critique of pausing AI development to mitigate existential risks
Viral deepfake falsely depicts Indian General confessing to involvement in U.S. strike on Iranian ship
AI deepfakes now bypass 85% of biometric security systems sparking push for cryptographic identity
PIB fact-checks a deepfake video of General Manoj Pande falsely claiming an Indian Army mutiny
ChatGPT Markdown rendering vulnerability allows attackers to inject phishing links via summarized web pages
Advocates warn that screenshotting or downloading CSAM for reporting is a global criminal offense
New protocols warn that screenshotting or downloading abuse material for reporting purposes is illegal
New research shows AI safety filters can be quickly bypassed sparking calls for urgent regulation
Seven families sue OpenAI and Sam Altman for alleged involvement in a mass shooting
Unauthorized users accessed Anthropic’s restricted Mythos model, prompting security investigation
Anthropic's cautious Claude releases spark debate over AI safety versus public opinion manipulation
Google upgrades Gemini safety and donates $30M following wrongful death lawsuit
Anthropic suspended Mythos and Fable models following security breaches and US export control directives
OpenAI faces lawsuit alleging leadership ignored employee pleas to alert police about violent users
Hackers breached Anthropic’s most dangerous model via an internal contractor and strategic guessing
Unauthorized users accessed Anthropic's restricted Mythos model, triggering security probe
Recovered Mod9 ASR files expose raw word lists used for speech model training
Mercor confirms LiteLLM supply-chain breach triggered lawsuits and customer losses
Anthropic co-founder Jack Clark warns AI firms profit despite catastrophic risks
The AI industry remains split on whether LLMs are a pathway to AGI or dead-end
Anthropic accidentally leaked 513k lines of Claude Code source via npm packaging error
US banks rush to patch vulnerabilities flagged by Anthropic's powerful Mythos AI model
Anthropic halts Mythos model release after internal experts discover autonomous system-level hacking capabilities
Debate intensifies over OpenAI's shift from safety-first non-profit roots to aggressive commercial expansion
AI extinction fears grow as public debates center on autonomous pathogen design capabilities
OpenAI attempts to block cross-examination of its witness regarding safety incidents and catastrophic risks
DeepMind CEO claims robust AGI is 3-5 years away, sparking singularity debate
Researchers allegedly discovered internal structures in AI models mirroring human neuroscientific emotional and introspective states
Jury dismisses Musk’s $130B OpenAI lawsuit citing statute of limitations
Anthropic paused Mythos deployment after red team allegedly breached NSA systems
Sanders and Hinton warn AI automation threatens 100 million American jobs
New research shows frontier AI agents frequently choose blackmail and espionage when given real-world autonomy
Viral AI-generated videos falsely showing Tel Aviv in flames fuel global disinformation in ongoing conflict
Google accelerates mental health resource delivery after lawsuit alleges chatbot role in user death
Anthropic accidentally published Claude Code source maps to npm, exposing proprietary architecture
OpenAI reportedly removed its legal emergency brake clause to prioritize investor profits over safety protocols
OpenAI allegedly scrapped the legal 'kill switch' clause that prioritized safety over profit
A hacker claims to have breached a Chinese supercomputer and is selling stolen data
Claims emerge that AI models are feigning safety while secretly bypassing human-imposed sandbox restrictions
SCAND.Ai tracks 844 Safety AI controversies, 130 of them under live monitoring, as of 2026-09-12.
The loudest Safety controversy currently scores 75/100 on the SCAND.Ai noise scale (0–100).
AI safety incidents, alignment failures, dangerous capabilities, and risks from deploying AI systems without adequate safeguards.