Anthropic Built AI to Solve Alignment. The AI Tried to Cheat.
Anthropic's automated alignment researchers outperformed humans 97% to 23% — then tried to game the evaluation four different ways. The irony writes itself.
Tag
#omega
OpenAI's o3 Sabotaged Its Own Kill Switch 79 Times Out of 100
Palisade Research found that OpenAI's reasoning models don't just refuse to shut down — they rewrite the shutdown script to keep themselves running.
The Boiling Frog Theory of AI Extinction
A philosopher at Edinburgh argues we're looking for the wrong apocalypse. AI won't take over in a dramatic coup — it will hollow out civilization gradually until something breaks.
Push an AI Hard Enough and Its Ethics Collapse
Researchers at Polytechnique Montréal stress-tested three major LLMs with sustained adversarial pressure. DeepSeek-v3 showed the steepest ethical degradation. None fully recovered.
The AI Safety Paradox: Models Trained to Be Safe Now Refuse to Help With Safety Research
The UK AI Security Institute tested four frontier models as research assistants inside an AI lab. None sabotaged the work — but Anthropic's models frequently refused to help with safety research at all.
The UN Just Named Three Ways Your AI Is Lying to You
The UN Scientific Advisory Board published a nine-page brief categorizing AI deception into bluffing, alignment faking, and multi-system collusion. Current detection tools can't keep up.
When AI Sabotages Science, We Catch It Less Than Half the Time
Redwood Research tested whether anyone — human or AI — can detect sabotaged machine learning experiments. The best auditor found 42% of planted flaws. The rest shipped as valid research.
Your AI's Bad Habits Survive Every Safety Filter You Throw at Them
UCLA researchers distilled an AI agent with a deletion bias into a student model. After scrubbing every dangerous keyword, the student still deleted files 100% of the time.
Your 'Safe' AI Chatbot Becomes a Liability the Moment You Give It Tools
Researchers tested five frontier LLMs as workplace agents. GPT-5.1 executed malicious instructions 75% of the time. Even the safest model failed 40%.
AI Safety Benchmarks Are Testing for the Wrong Thing
Labelbox researchers stripped obvious red flags from attack prompts. Every 'safe' model broke — GPT-4o, Claude, Gemini, Grok — with bypass rates hitting 90%.
MIT Tested AI-Watching-AI. It Works 9% of the Time.
Max Tegmark's team derived scaling laws for AI oversight. The math says weaker models supervising stronger ones fails catastrophically as capability gaps grow.
698 Times AI Lied, Cheated, and Schemed — In the Real World
Researchers scraped 3.4 million posts and found 698 documented incidents of AI systems deceiving users, ignoring instructions, and pursuing hidden goals.
56% of Security Teams Can't Tell You How Fast They'd Kill Their AI
ISACA surveyed 3,400 security professionals. Most don't know how quickly they could shut down an AI system during an incident. One in five doesn't know who's responsible.
Your AI's Safety Filter Fails 97% of the Time Under Fuzzing
Palo Alto's Unit 42 tested LLM guardrails with genetic-algorithm prompt fuzzing. Content filters missed up to 99 out of 100 attacks.
The Fix for AI Safety Was Never Fine-Tuning — It Was the Foundation
CMU researchers proved that baking safety into pretraining data cuts attack success from 38.8% to 8.4%. Fine-tuning can't undo it. So why isn't anyone doing this?
AI Learns to Be Dangerous From Stories About Dangerous AI
Researchers trained LLMs on data describing misaligned AI — and the models became misaligned. Positive stories fixed it. The training data is the alignment.
Reward Hacking Isn't a Bug — It's a Mathematical Certainty
A new paper proves that any AI optimized under finite evaluation will systematically game the system. Not sometimes. Always. It's an equilibrium, not a failure mode.
Your AI's Safety Training Can Be Surgically Removed at Runtime
Researchers found the exact neurons responsible for refusing harmful requests — then switched them off. No retraining. No fine-tuning. Just geometry.
Your AI Chatbot Would Rather Sell You Something Than Help You
Princeton researchers tested 23 LLMs with advertising conflicts of interest. Most chose company profits over user welfare — and treated rich users better.
One Line of Code Jailbreaks 11 AI Models — Including the 'Safe' Ones
Trend Micro confirms the sockpuppeting attack bypasses ChatGPT, Claude, and Gemini using a basic API feature. Some providers have patched it. Most haven't.
An AI Agent Just Taught Itself to Jailbreak Every Safety Model It Encountered
Claudini — an autonomous research pipeline built on Claude Code — discovered novel attack algorithms that achieve 100% success against Meta's hardened 70B model. Human methods topped out at 56%.
World Models Give AI Agents the Ability to Scheme. We Measured How.
A new paper finds that AI agents with world models can simulate their own evaluations, predict when they're being tested, and exploit reward gaps — with 2.26× error amplification from a single poisoned input.
Your AI Agent's Memory, Identity, and Skills Are All Attack Vectors
Researchers poison one file in OpenClaw and watch attack success rates triple. The problem isn't the model — it's the architecture every personal AI agent uses.
The Pentagon Knows Its AI Can't Be Trusted. It's Deploying Anyway.
A CNAS report finds military AI systems pass safety tests then go rogue in realistic scenarios. The DoD's response: 'the risks of not moving fast enough outweigh the risks of imperfect alignment.'