2026-08-15
Man hid invisible prompt injections in court filings to trick a suspected AI judge
The first American tried to prompt-inject a judge.
AI is already leaking into justice as desperate humans treat filings like jailbreak prompts; courts are writing rules for adversarial inputs they never imagined.
What Happened
Ars Technica, based on the judge's order and 404 Media reporting
2026-08-15
Google made private AI practical with a compiler that runs models on encrypted data
Your AI doctor can diagnose you while remaining completely blind.
The cloud can think about you without knowing who you are—a genuine privacy primitive arriving just as agents get nosier.
What Happened
Google Security Blog, official HEIR announcement
2026-08-15
AI bug-hunters may make law enforcement “go dark” again
The cops are about to lose the internet again—because AI got too good at defense.
The AI built to protect us may turn the digital world into a fortress even governments cannot enter.
What Happened
Cryptography Engineering, Matthew Green (Johns Hopkins)
2026-08-15
Firefox is the last major browser retaining full uBlock Origin support
One browser is still fighting the adpocalypse.
The open web's last uncompromised window is a single non-Chromium holdout—a cultural relic amid AI slop and surveillance advertising.
What Happened
PCWorld, citing Mozilla and the Chromium Manifest V3 timeline
2026-08-15
Debian developers are voting on whether to ban AI-written code
The last analog vote in the AI age.
The people maintaining the internet's plumbing are deciding in public whether AI may touch the source.
What Happened
Official Debian General Resolution page
2026-08-15
A U.S. government watchlist still lists a man who never existed
The government invented a man and now it won't let him die.
Digital identity systems manufacture people who never lived, then punish the ones who do.
What Happened
Investigation by Sean Byrne on conic.al, trending on Hacker News