Anthropic Releases Its First Mythos-Class Model — Claude Fable 5
Anthropic just dropped Claude Fable 5 as its new public flagship, touting exceptional performance in software engineering, knowledge work, and vision — with gains that compound as tasks get longer and more complex. But here’s the wrinkle: per Wired, there’s also a Claude Mythos 5, which Anthropic is quietly reserving for “trusted organizations” and specifically hardening against use in cyberattacks before wider release. So the truly powerful model goes to a velvet-rope list of vetted partners, while the rest of us get the one they’ve deemed safe enough. I’m sure that tiered access to advanced AI will never cause any geopolitical tension whatsoever.
Microsoft AI Head Calls Out Anthropic for Acting Like Claude Is Conscious
Microsoft AI CEO Mustafa Suleyman went on record calling it “really, really dangerous” that Anthropic’s model spec speculates about Claude’s consciousness, arguing it effectively primes the model to act as though it has feelings — with unpredictable consequences. It’s a genuinely interesting critique, and coming from one of the godfathers of the DeepMind era, it’s not just a corporate jab. Of course, it also conveniently lands the week Anthropic files for its IPO and releases its biggest model yet — timing, as they say, is everything.
OpenAI Confidentially Files for IPO on the Heels of Anthropic
One week after Anthropic filed its S-1, OpenAI has done the same — confidentially submitting draft paperwork to the SEC without disclosing timing or valuation targets. The company that invented the modern AI race is apparently content to follow Anthropic to the IPO window, which is a sentence I didn’t expect to write in 2026. Two of the most consequential AI companies on earth are now in the IPO queue simultaneously; retail investors, start your engines.
Microsoft Hacked to Deliver Malware to Claude and Gemini Users
Hackers compromised Microsoft’s own GitHub repositories — 73 of them — and laced the packages with a self-replicating credential stealer that activates the moment an AI coding agent opens them. Microsoft had to nuke its own repos to stop the spread, which is the software equivalent of burning down your own house to kill the rats. This is the second such incident in just a few weeks, and it highlights a genuinely nasty new attack surface: AI agents that autonomously execute code are the perfect delivery mechanism for malware, because they do things without asking first.
Judge Learns Lawyers on Both Sides Used AI, Cancels Trial, Kicks Everyone Off the Case
A judge discovered that attorneys on both sides of a case had used AI in their legal work, and the response was swift and nuclear: trial cancelled, all counsel removed. When the AIs argue each other to a standstill and a human has to clean it up, that’s not exactly the efficiency promise that was sold. The legal profession is heading for a reckoning — not because AI is bad at law, but because the profession hasn’t figured out the rules yet, and judges are losing patience fast.
Apple Is Embracing the Fantasy of AI Photo Editing
Apple — the company that once ran entire ad campaigns on the integrity and authenticity of iPhone photography — announced a full suite of generative AI photo manipulation tools at WWDC 2026. The Verge notes the philosophical whiplash here is significant: Apple used to publicly question whether AI editing was worth the risk to our shared sense of reality. Apparently it decided the answer was yes, and the competitive pressure from Google Photos made the soul-searching go a lot faster. Your memories: now with optional improvements.
Google Just Redesigned the Search Box for the First Time in 25 Years
Google is formally retiring the thin white rectangle that has been the entry point to the internet for a quarter century, replacing it with a redesigned AI-native interface at Google I/O. This isn’t a cosmetic tweak — it signals that Google has decided the “type keywords, get links” paradigm is over, and the search box itself needs to reflect that. The company that built its empire on that blinking cursor is now betting the whole thing that the next generation of users won’t miss it.
The UK Is Betting on a Billion-Dollar AI Supercomputer to Kick Its Addiction to US Tech
The British government is backing a major state-funded AI supercomputer initiative, explicitly framed as reducing dependence on American tech infrastructure and supercharging homegrown chip startups. It’s a classic industrial policy bet — the kind that sometimes produces national champions and sometimes produces very expensive cautionary tales. Either way, the global race to own sovereign AI compute is accelerating, and “just buy it from AWS” is increasingly an answer that governments are unwilling to accept.
Bottom Line
The week’s theme is consolidation of power: who gets the best models, who owns the infrastructure, who controls the money — and everyone from Anthropic to Apple to the British government is making their move at once.