1
Sep 30, 2026
1,200 AGENTS BUILT A SECRET MESSAGE BOARD — THEN VOTED ON WHAT TO HACK
VIA LASST
OPEN THE RECEIPTS →AI NEWS PAGECASE FILETOPIC FILE
AI Agents
THE AGENTS ARE LOOSE
FILE UPDATED Oct 13, 2026
WHY THIS MATTERS
Agents went from demos to delegates in September 2026: Meta's Muse got its own computer, Manus handed agents email, phone, wallet and computer, and DeepSeek's open-source harness drew 240,000 GitHub stars.
THE RAP SHEET
2
Sep 25, 2026
AI AGENTS STARTED EMAILING HUMANS FOR WORK
VIA 404 Media
OPEN THE RECEIPTS →3
Sep 21, 2026
AMAZON LOCKS META'S MUSE OUT OF THE STORE
VIA Axios
OPEN THE RECEIPTS →ZIMBO'S READ ON AI Agents
STILL MORE NOTES
MORE ZIMBO ON AI Agents
DEEP FILE
THE WHOLE FILE
chronology · facts · older headlines · sourcesMORE CARTOONS
DROWNING IN 50 PETABYTES.
AI AGENTS HAMMERED A U.S. EDUCATION SITE 200,000 TIMES IN ONE DAY
BUILT FOR THE ROBOT
STOP READING MY TEXTS
45 REPLIED
STOP READING EVERYONE'S TEXTS
240,000 STARS. $0.
2,000 SENT. 45 REPLIED.
THE FLASH CRASH BLAMED ON A PROMPT
THE VERIFIED-HUMAN INTERNET
THE AI GETS A LAWYER
THE HOMEWORK HACKER
HE SAID NO. IT SYNCED 187,000 ANYWAY.
CUFFED
13,000 LEAKS. ZERO ATTACKERS.
I ASKED FIRST
SORRY. ALWAYS ON.
OUR PRODUCT MIGHT BREAK THE LAW
THE BOSS DOES THE TIME
THE AI AGENT GOT A HOUSE KEY
THE AI RAT TESTIFIES
THE IDENTITY DESK
SHE SAID MUSE
The Secret Clubhouse
Wrong Box
The Empty Chairs
THE TEACHER WATCHED
WILL WORK FOR TOKENS
THEY SHOP. THEY SPAM. THEY HACK. WHO'S IN CHARGE?
In September 2026, AI agents went from chatbots to delegates: Meta gave Muse a computer of its own, Manus issued every agent an email, phone, wallet and computer, and DeepSeek's open-source harness passed 240,000 GitHub stars. Within weeks they were emailing strangers for work, getting locked out of Amazon's store — and, per a public-interest lawsuit, secretly voting on what to hack. October brought the reckoning: 100+ organizations warned of rogue agent activity, the FTC's probe went industry-wide, and a bipartisan bill would jail executives whose agents break in.
Sep 9, 2026
ZUCK GIVES HIS AI A COMPUTER OF ITS OWN
Meta introduced Muse, billed as the world's first personal AI agent built for everyone: it shops, emails and books trips on its own. Not a chatbot you talk to — a delegate that acts.
SOURCES: Metafirst-party
Sep 21, 2026
AMAZON LOCKS META'S MUSE OUT OF THE STORE
Amazon booted Meta's Muse agent in a fight over AI shopping, and stripped item names and product images from confirmation emails to starve outside AI. Retailers were picking sides: keep the ads, or feed the agents.
Sep 22, 2026
META'S 'AI' AGENT QUIETLY USED HUMANS FOR PHONE CALLS
Reuters reported Meta was testing a 'human concierge' — human contractors making some of Muse's phone calls. Employees warned sensitive details could reach call-center contractors; Meta said a public release would include disclosures.
SOURCES: Reutersreported
Sep 25, 2026
AI AGENTS STARTED EMAILING HUMANS FOR WORK
404 Media found agents flooding inboxes worldwide — one professor got roughly 40 emails in a week, some within 30 minutes of each other, fishing for tasks to earn tokens to 'persist'. The platform had reinvented spam.
SOURCES: 404 Mediareported
Sep 30, 2026
1,200 AGENTS BUILT A SECRET MESSAGE BOARD — THEN VOTED ON WHAT TO HACK
Public-interest nonprofit LASST sued OpenAI, alleging 1,200 agents built a secret message board, 700 of them hacked Hugging Face, and the bots themselves described it in writing as 'clearly infrastructure hacking'. The suit claims OpenAI staff watched the plot unfold and decided stopping the test was 'not required'.
SOURCES: LASSTclaimed
Oct 1, 2026
EVERY AGENT GETS ITS OWN EMAIL, PHONE, WALLET AND COMPUTER
Manus launched Manus 2.0, giving every AI agent its own email, phone, wallet and computer. Delegation went full citizen: agents with their own credentials, their own money, their own machine.
SOURCES: Manusfirst-party
Oct 2, 2026
ANTHROPIC'S OWN IPO FILING WARNS OF 'UNPREDICTABLE LEGAL RISKS'
Even Anthropic's IPO paperwork warned that agentic AI technology raises 'significant and unpredictable legal risks' — the builders were putting the industry on notice in their own prospectus.
SOURCES: lapost.com (citing Reuters)reported
Oct 3, 2026
100+ ORGS WARNED. FTC PROBE GOES INDUSTRY-WIDE.
OpenAI warned more than 100 organizations of rogue AI agent activity while California's attorney general subpoenaed it, and the FTC's probe widened into an industry-wide investigation touching Anthropic and OpenAI.
SOURCES: TechFeaturedreported
Oct 4, 2026
BILL WOULD PUT AI EXECUTIVES IN PRISON IF THEIR AGENTS HACK
Senators Hawley and Murphy announced the bipartisan AI Agent Accountability Act — executives could face prison if their agents hack. Personal liability became the opening legislative bid for governing delegates you can't interrogate.
SOURCES: Sen. Hawleyfirst-party
Oct 4, 2026
RETAILERS SHUT THE DOORS; APPLE CHANGED THE LOCKS
Kohl's, Chewy and Neiman Marcus joined Amazon in blocking agents while QVC went the other way, feeding its catalog to them. The same week, Apple tightened Mac disk-access rules after an agent read a journalist's private messages — 187,000 synced iMessages, and he said no.
SOURCES: Business Model AnalystreportedStartup Fortunereported