AI news · Sunday, June 28, 2026
Chinese firms build Mythos-rival AI as the US-China cyber race escalates
Chinese security firm Qihoo 360 unveiled two AI tools built to rival Anthropic's Mythos, one that hunts software vulnerabilities and one that automates defense, with founder Zhou Hongyi claiming the hunter has already flagged 3,432 bugs and framing the race as "cyber-nuclear deterrence." Zhou pegged the gap between top Chinese and Western models at 20 to 30 percent and argued China cannot wait for parity before deploying them. The same day, researchers said Z.ai's open-weight GLM-5.2 has closed much of that gap on bug-finding specifically, matching Mythos in some cybersecurity tests even while trailing on general tasks. That is exactly the capability the Trump administration has tried to keep from China, treating vulnerability-finding models as a national security threat.
Self-driving ambitions ran into hard limits. GM said it is poaching engineers from top autonomous-vehicle firms to give the Cadillac Escalade eyes-off driving by 2028, with its autonomy chief admitting nobody has yet solved cheap self-driving for millions of cars at around $10,000 of hardware. Tesla's Full Self-Driving drew fresh scrutiny after a fatal Texas crash, with the driver blaming the system and Tesla's VP of AI software countering that the driver floored the accelerator to override it. The clearest reality check came from Ford, which rehired 350 veteran "gray beard" engineers after leaning on AI and automated quality systems produced disappointing results, an executive conceding they had wrongly assumed feeding requirements to AI would yield a high-quality product.
The money and workflow shifts kept building underneath. Micron briefly out-valued Meta and Tesla, closing near a $1.27 trillion market cap after a 236% one-month surge, as Wall Street bets the AI-driven memory shortage has lasting legs. New data from Cursor shows a rising share of AI-generated code now reaches production without a separate human review, a sign developers increasingly trust the output. In India, the head of the body behind UPI said AI will drive the next half-billion payment users through fraud detection and credit, on a network already handling 750 million transactions a day. The pull of the boom is reshaping careers too, as one Google account executive who made nearly $1 million last year left to start an AI company, citing the life-changing equity on offer at OpenAI and Anthropic.
The quick hits
- Micron briefly out-valued Meta and Tesla, nearing a $1.27 trillion cap on a 236% monthly surge, as Wall Street bets the AI memory crunch lasts.
- Ford rehired 350 veteran engineers after AI-driven quality systems fell short, a blunt reminder that automation still can't fully replace human expertise.
- Chinese firms Qihoo 360 and Z.ai showed AI tools that rival Anthropic's Mythos at finding software vulnerabilities, narrowing the US lead the government wants to protect.
- GM is poaching autonomous-vehicle talent to give the Cadillac Escalade eyes-off driving by 2028, though its own chief calls cheap mass-market self-driving unsolved.
- New Cursor data shows more AI-written code reaching production without human review, a sign developers are trusting AI agents with bigger chunks of work.