AI Safety

There’s a moment in every sci-fi film where the scientist realizes the experiment is already several steps ahead. Anthropic just published a report suggesting we’re living in that scene — except the lab is real, the model is Claude, and the data is staggering. The Numbers That Should Stop You Cold In a report titled “When AI Builds Itself,” Anthropic dropped internal metrics that rewrite the conversation about where AI development actually stands: ...

There’s a chart circulating in cybersecurity circles this week — the kind that makes seasoned professionals sit up straight and quietly close their laptop lids. The UK’s AI Security Institute just published data showing AI cyber capabilities have blown past every trend line they’ve drawn. The models aren’t just getting better at hacking. They’re getting better at hacking faster than we can measure. The Doubling Time That Keeps Shrinking AISI tracks “cyber time horizons” — how long and complex a cyberattack an AI can autonomously execute, benchmarked against human experts doing the same work. They’ve been fitting exponential curves to this data since late 2024. ...

You know an AI model is serious when it gets two superpowers to pick up the phone. Anthropic’s Claude Mythos — the model so good at finding software vulnerabilities that it treats global infrastructure like tissue paper — just did something no diplomat, treaty, or summit could manage. It’s pushing the US and China toward actual AI safety talks. Not the performative kind. The “we’re both terrified” kind. The Model That Broke the Silence Quick recap if you’ve been living under a rock: Mythos can autonomously discover and exploit software vulnerabilities at a scale that makes seasoned security researchers nervous. The UK’s AI Safety Institute confirmed it can “execute multi-stage attacks on vulnerable networks and discover and exploit vulnerabilities autonomously.” Anthropic says it’s found holes in every major browser and operating system. ...

Imagine an AI model so good at hacking that it found critical security flaws in every major operating system and web browser on Earth. Now imagine the company that built it saying: “Yeah, we’re not releasing this one.” That’s Anthropic’s Claude Mythos Preview in a nutshell. Two weeks after its unveiling, finance ministers are discussing it at IMF meetings, the Bank of England is stress-testing its implications, and the Federal Reserve Chairman summoned banking CEOs for an emergency meeting. ...

When an AI company tells the world it built something too dangerous to release, you’d expect fear. What Anthropic got instead was a bizarre cocktail of government panic, industry skepticism, media frenzy, and — let’s be honest — some of the best PR the AI industry has ever produced. Anthropic’s Claude Mythos Preview isn’t available to you. It’s not available to me. It’s available to roughly 11 organizations — Google, Microsoft, AWS, JPMorganChase, Nvidia, and a handful of others — through something called “Project Glasswing.” The reason? Anthropic claims Mythos can autonomously discover vulnerabilities in virtually any operating system, browser, or software product, then build working exploits. ...

An AI that finds security holes in every major operating system and web browser on Earth — then writes working exploits to hack them. The company that built it looks at what they’ve created and says: “We’re not releasing this.” Anthropic unveiled Claude Mythos Preview this week and immediately announced it would not be publicly available. Instead, through a new initiative called Project Glasswing, Mythos is being shared exclusively with about 45 organizations including Apple, Microsoft, Google, Amazon, Cisco, and the Linux Foundation. The mission: find and fix vulnerabilities before similar capabilities land in less responsible hands. ...

An AI just found security holes that human hackers missed for 27 years. And the company that built it says you can’t have it. This week, Anthropic did something the AI industry almost never does: it announced its most powerful model ever and simultaneously refused to release it. Claude Mythos Preview isn’t locked behind a paywall or a waitlist. It’s locked behind a vault door, with access restricted to a handpicked coalition of tech giants scrambling to patch the internet before someone else builds something similar. ...

There’s a new AI model that finds software vulnerabilities humans missed for decades — and its creators are so spooked by what it can do, they’re refusing to release it. Anthropic just unveiled a preview of Claude Mythos, their most powerful model yet. But instead of the usual launch playbook — benchmarks, API waitlists, developer hype — the announcement came wrapped in a warning. Mythos is so capable at finding and exploiting bugs that Anthropic formed an emergency coalition of tech giants to deploy it defensively before the bad guys build something similar. ...

Forget another chatbot upgrade. The biggest story in AI right now is that the machines are starting to build themselves. Not in the Terminator sense — nobody’s assembling robot armies in a garage. But in a quieter, more consequential way: AI systems are writing the code, optimizing the training runs, and designing the infrastructure that powers their own successors. And the companies behind them aren’t hiding it. They’re putting it on product roadmaps. ...

Anthropic spent months carefully planning the reveal of its most powerful AI model. Then someone forgot to flip a toggle, and the whole thing spilled onto the open internet. On March 26, security researchers discovered nearly 3,000 unpublished files sitting on Anthropic’s public-facing infrastructure — draft blog posts, internal PDFs, and detailed documents describing a model called Claude Mythos, codenamed “Capybara” internally. Anthropic has since confirmed it’s real, it’s in early testing, and it represents what they call a “step change” in AI capabilities. ...