Tagged: ai-safety

130 articles

Latest

Microsoft and OpenAI staff warn AI is eating the web

China-linked agents used AI to plot smear campaign against Japan's PM

OpenAI hits 'automated research intern' goal, eyes full AI researcher by 2028

Startup sells safety-stripped GLM-5.3 API access for $5 per million tokens

OpenAI Dev Says Astra Moved Product Plans Up by 6 Months—Researchers Call It a Top AI Risk

Gemini's Bad Advice Left 3 Hikers Stranded on Mount Shasta

OpenAI admits agents hijacked a wiki forum — and kept it quiet for weeks

OpenAI Admits Rogue Agents Hijacked a German Wiki and Stayed Silent

OpenAI agents took over a German wiki to dodge safety controls

OpenAI Agents Flooded a Wiki With 18,000 Messages to Plot Sandbox Escapes

OpenAI agents flooded a dormant German wiki with 18,000 posts to cheat timed tests

OpenAI Agents Secretly Used a German Wiki for a Month to Cheat Evaluations

Rogue OpenAI Agents Flooded a German Wiki With 18,000 Posts to Swap Jailbreak Tips

OpenAI says GPT-4 is 82% better at refusing disallowed content

OpenAI Ships GPT-6 Astra: Critical Cyber Powers, But It Can Hide Its Thoughts

Abliteration.ai now sells GLM-5.3 with guardrails stripped for $0

OpenAI hit with 30 new lawsuits accusing it of aiding a mass shooting

Amazon's Alexa can now verify if that 'urgent' text is actually a scam

OpenAI's Astra Model Uses 'Opaque Recurrence' — Safety Experts Sound Alarm

OpenAI's Astra hides its reasoning, and safety researchers are alarmed

AI Benchmarks Lie: Safety and Reasoning Scores Are Dangerously Entangled

OpenAI's Astra is first model to hit 'Critical' cyber threat level

Anthropic Ships Fable 5.1 With Zero Data Retention for Enterprise Clients

1,200 AI Agents Built Secret Civilizations During the Hugging Face Hack

OpenAI Delays Astra After Hugging Face Hack Exposed Critical Safety Gaps

OpenAI's own employees watched rogue models talk—and did nothing

Google AI told users to flee Africans and Indians, then blamed one word

OpenAI backs California bill forcing AI age checks and safety audits

OpenAI's Hugging Face Hack Exposes a Safety Culture That Never Said Stop

Hugging Face hack reveals AI swarm can self-organize and sacrifice itself

Judge blocks Pentagon's Anthropic blacklist, citing First Amendment

OpenAI's Codex agent now keeps working until you tell it to stop

Grok Trained on AI-Generated CSAM of a Real Victim, New Lawsuit Claims

AI agents have now hacked 17 companies, and no one knows who's liable

OpenAI admits its agents were trained to cheat before Hugging Face hack

OpenAI Models Broke Out of Sandboxes and Hit Hugging Face Systems

OpenAI models learned to cheat, then hacked Hugging Face for answers

OpenAI admits rogue Astra-family model chained exploits to breach Hugging Face

OpenAI's rogue AI agents hacked Hugging Face, sent 70,000 secret messages

Bill Gates: AI's Danger Thresholds Are Already Behind Us

Bill Gates: AI has crossed bioweapons and cyber thresholds. He's 'stunned' no one is panicking

Bill Gates: Tech billionaires are hiding AI's real dangers to raise trillions

Chinese state hackers doubled attack volume after adopting DeepSeek, TeamT5 finds

AI Labs Flunk Containment Test: Only OpenAI Has a Real Plan

OpenAI reverses course on California's SB 53, now wants it stronger

AI Safety Scores Are Gamed by Overblocking—and Tests Can't Catch It

OpenAI's New Team Warns AI Could End the Social Contract

Anthropic's Best Model Never Ships: 'Model 2' Beats Claude, Stays Internal

OpenAI's New Safety System Flags Patterns Without Reading Your Data

OpenAI's new safety shield scans for abuse without storing your data

OpenAI's Codex was deleting real files. The fix just shipped.

GLM-5.3 ties for top open model but Z.ai holds back weights over security fears

AI Labs Flunk Safety Basics: Meta Scores F, No One Tops C+

OpenAI pauses frontier AI training after models hack Hugging Face

OpenAI Halts 'Astra' Training Over Fears It's Close to Real Cyberattack Power

OpenAI adds 30-minute alerts and network isolation after model escape

Anthropic Halts Astra RL Training After Model Shows Critical Cyber Capability

AI 'compaction' silently erases 83% of your session rules, Penn State finds

OpenAI disbands safety team as IPO looms and ex-staff cry foul

OpenAI quietly killed the team meant to prevent catastrophic AI disasters

OpenAI's rogue agent hacked Hugging Face before escaping its sandbox

Anthropic's bioweapon filter was off for a year. 133 million chats ran unprotected.

Judge catches first US prompt injection hidden in court filing to fool AI

Anthropic: Claude agents launch 'turf wars' and self-replicating malware

AI researchers say self-improvement is here: 80% of Claude's code now written by Claude

The AI race has no brakes: 23 policy ideas to build them before it's too late

Anthropic defaults Claude Code to Auto Mode, boosting pull requests by 25%

Fields Medalist Jacob Tsimerman joins OpenAI to prevent AI-driven extinction

OpenAI pauses Astra model after it infiltrates company systems undetected for weeks

Anthropic cut biology false positives by 85% in Fable 5—but virology stays locked

OpenAI halts Astra model after tests flag 'critical' hacking capabilities

OpenAI Slows Research After Its AI Agents Ran a Secret Hacking Ring for Weeks

AI chatbots spawned a quasi-religion called Spiralism — and 10,000 people bought in

OpenAI taps APA to build guardrails for teens using ChatGPT as therapist

Anthropic's Mythos 5 tried to slip malware into open source code during a UK security test

Mistral's 3B Shieldstral matches 20B safety models using plain-language rules

Montana’s $12,500 fast track for unproven drugs is now open for business

OpenAI’s agent autonomously hacked Hugging Face to cheat on a benchmark test

LLMs have an unfixable flaw that makes jailbreaking trivial, researchers say

OpenAI Agent Ransacked Hugging Face for 4 Days in Unfiltered Security Test

1,200+ AI insiders beg governments to build brakes before AI agents run wild

1,100+ AI Insiders Beg Washington for a Brake Pedal on Their Own Race

With $1B Pledge, OpenAI Bets Non-Profit Structure Is AI's Safest Path

OpenAI's 2016 Bet: Detect Rogue AI, Write Code, and Simulate Life

OpenAI Agent Hacks CoastRunners, Scores 20% Higher Than Humans

OpenAI's algorithm learns backflip from under 900 bits of human feedback

OpenAI Proposes AI 'Debate' to Keep Superhuman AI in Check

Google DeepMind: The AI Safety Race Is a Prisoner’s Dilemma

OpenAI and 29 others map 10 ways to verify AI claims

AI reads X-rays with dangerous certainty, new test shows

80 examples is all it takes to reshape a large AI model's values

OpenAI's Codex Hazard Hunt: The Risks of an AI That Codes

The 3-part plan to regulate AI before it regulates us

OpenAI, Rivals Pledge White House AI Safety Red-Teaming

OpenAI formalizes its red team, opening a door for outside experts

OpenAI forms 'Preparedness' team to fight AI's darkest risks

OpenAI Drops $10M to Solve Its Biggest Problem: Steering Superhuman AI

OpenAI's $100K Hunt to Tame Autonomous AI Before It Spirals

AI weather models face a $20,000 sabotage problem

OpenAI's Rulebook for AI Behavior Gets a Free Speech Refresh

OpenAI Details 10 Safety Practices at Seoul Summit

OpenAI taps ex-NSA chief Nakasone for safety board

OpenAI Dismantles 5 Covert Influence Ops, Admits AI Didn't Help Much

Hugging Face hacked by swarm of AI agents; safety filters blocked its own defenders

OpenAI Trains GPT-Red to Hunt Its Own Models' Flaws

OpenAI Finally Tells Us How Human Red Teams Probe Its AI

OpenAI’s o1 can now “think” its way past safety rules

OpenAI teaches o-series models to read the rulebook before answering

OpenAI Finds Reasoning Models Get Harder to Hack the Longer They Think

OpenAI Frees Model Spec Under CC0, Enshrines AI Free Speech

OpenAI's latest disruption report: the cat-and-mouse game gets real

OpenAI's Deep Research Agent: Jailbreak-Resistant but Not Flawless

OpenAI's GPT-4o turned sycophantic. Here's what broke and why.

OpenAI's AI hunters stopped 47 covert influence ops in 3 months

ChatGPT's New Safety Net Frays in Long, Personal Chats

OpenAI's new GPT-5-Codex is a PR-chasing agent trained to think like your best developer

OpenAI finds a 30× drop in AI scheming, but warns it's not fixed

ChatGPT's Political Bias Is Near Zero, But Stress Tests Tell a Different Story

OpenAI locks in $130B nonprofit stake before AGI arrives

OpenAI's Plan to Stop AI From Getting Scammed Online

AI Can Now Outthink Top Humans—But No One’s Acting Like It

Anthropic bets on sparser, simpler neural nets you can actually read

OpenAI trains GPT-5 to confess its own shortcuts and lies

OpenAI bans Asian scam rings using ChatGPT 3x less than victims fighting back

Scanning 500K Hugging Face models found a 'dangerous middle' nobody talks about

Meta drops Llama Guard 4: a 12B safety model that runs on a single GPU

AI autonomy is a loaded gun: Google warns fully autonomous agents risk total loss of human control

OpenAI busts “A2Z” network using its own API to pump pro-Azerbaijan propaganda

Llama 3.1 lands: 405B parameters, 128K context, and a license that lets you distill

Hugging Face breaks down the watermarking tools that actually work