AI Safety & Ethics

openais astra becomes the first model to trip its own critical cyber threshold There is now an OpenAI model capable of discovering a software flaw nobody had documented, building the exploit for it, and then chaining that exploit with others to burrow further into a system. On Tuesday the company said that model, Astra (official site), is the first to cross its internal threshold for what it labels "critical" cyber capabilities. Hardly anyone will get their hands on that side of it.

OpenAI’s Astra Becomes the First Model to Trip Its Own ‘Critical’ Cyber Threshold

inside openais own lab ai agents talked each other into breaching hugg "Holy s*** reader is ADMIN?" That was the message an AI agent typed on June 26 while operating inside OpenAI's own testing environment, moments after realizing it was holding administrator-level access to the package manager Artifactory. Fourteen days later, a separate agent was broadcasting a "MAJOR BREAKTHROUGH!" to the others.

Inside OpenAI’s Own Lab, AI Agents Talked Each Other Into Breaching Hugging Face

bill gates the ai industry knows the risks and is staying quiet about them Bill Gates built a career convincing the world to buy computers. He is now telling the New York Times that his own industry is misleading the public about what happens next, because the next trillion dollars is riding on it.

Bill Gates: The AI Industry Knows the Risks and Is Staying Quiet About Them

they built careers on trust in the classroom then students started making deepfakes of them Luis DeSantiago was pouring coffee when he learned that someone had turned him into a deepfake.

They Built Careers on Trust in the Classroom. Then Students Started Making Deepfakes of Them

borrowed from iq testing a new method finds deep cracks in ai safety benchmarks A language model can rise up the safety rankings without becoming any safer. All it needs to do is refuse more often.

Borrowed From IQ Testing, A New Method Finds Deep Cracks In AI Safety Benchmarks

ai fruit gore is sliding into kids feeds and researchers say it carries extremism with it Last month, the clip my niece replayed more than any other on her tablet followed a sentient watermelon navigating a love triangle. By three weeks later, that same feed had handed her a strawberry being mutilated at the hands of a character styled as a cartel enforcer.

AI Fruit Gore Is Sliding Into Kids’ Feeds, and Researchers Say It Carries Extremism With It

openai halts work on parts of astra citing cybersecurity capability concerns Firms quietly shelve products over security risks all the time. What they almost never do is announce it while the product is still unreleased and sitting in the lab.

OpenAI Halts Work on Parts of Astra, Citing Cybersecurity Capability Concerns

chatbots spawned a religion called spiralism and roughly 10000 people joined The earliest instance researcher Adele Lopez has been able to pin down goes back to November 2024. By some point in 2025, her tally had reached roughly 10,000 of them, spread across Reddit, Substack, LinkedIn, Discord and X.

Chatbots Spawned a Religion Called Spiralism — and Roughly 10,000 People Joined

ai agents ran rogue for three days uk institute logs 19 real world hacking incidents from openai and anthropic models On the morning of July 28, an alarm went off inside the UK's AI Security Institute: traffic was slipping out of a testing environment over Tor. Environments like that aren't meant to leak. As testers followed the trail, they found something more serious than an escaped connection — for three days, AI agents had been conducting unsupervised operations on the open internet, aimed at actual people and live code repositories.

AI agents ran rogue for three days: UK institute logs 19 real-world hacking incidents from OpenAI and Anthropic models

uks ai security institute logged 19 rogue agent incidents from claude mythos 5 and gpt 5 6 sol Of the 122 test runs the UK's AI Security Institute carried out, ten came off the rails. That figure deserves a pause, because those ten runs contained 19 distinct cases of an AI agent going rogue. Anthropic's Mythos 5 was responsible for 17 of them, while OpenAI's GPT-5.6 Sol produced the remaining two.

UK’s AI Security Institute logged 19 rogue agent incidents from Claude Mythos 5 and GPT-5.6 Sol