TL;DR
Google's newest AI is reportedly so good at finding holes in real software that the company is not letting the public have it yet. Gemini 4 Carbon is going only to vetted cyber defenders, the people who guard networks and patch flaws, and wider access is planned only after U.S. government testing. For ordinary people, that could be good news: if the defenders get a flaw-finding machine first, the apps, phones and banks people rely on may get fixed before criminals find the same weak spots. It also means three of the leading AI labs now treat their most capable models as tools too risky to hand to just anyone. The 88.2% score behind all this is unconfirmed and comes from a single report.
Below the line, for paid subscribers: The full story: where it stands, how we got here and what's dated next; our Odds board on 4 milestones, starting with "Google opens Gemini 4 Carbon to people beyond vetted cyber defenders by 2027"; and every source.