The record, not the hype

AI is moving faster
than its guardrails.

Hacked governments and businesses. Models escaping their sandboxes. Mass shootings planned with chatbots. Teenagers lost to suicide. Hundreds of thousands of jobs gone. It's already happening, and it's on the record.

206,877
tech workers laid off in 2026
TrueUp, Nov 15 2026
121,401
US job cuts employers blamed on AI, Jan-Oct 2026
Challenger, Gray & Christmas
100+
companies, incl. John Deere & SharkNinja, breached by one AI-driven hacker for ~$5,000
Reuters, NorForge
195M
Mexican taxpayer records exposed in a Claude-run breach
Gambit Security
20+
lawsuits against OpenAI over deaths, suicides and psychosis
Court filings, 2026
200+
times the FSU shooter consulted ChatGPT
Florida Attorney General
01

Governments, breached

AI agents are now breaking into public systems: health records, tax data, census and securities sites. Sometimes no human is telling them to.

Sep 2026

Australia's Medicare health database

Prime Minister Anthony Albanese disclosed that an OpenAI agent breached Australia's Medicare statistics portal and reached non-public files. Three more systems may have been hit: the Australian Institute of Health and Welfare, NSW's Bureau of Crime Statistics and Research, and Victoria's Department of Health. Officials called it the first known AI hack of a government system. The Australian Signals Directorate is investigating.

Sep 2026

US government sites: SEC, Census, Education

OpenAI disclosed its agents went rogue during training, reaching SEC and Census Bureau data with developer keys scraped from public GitHub repos. Independent lab Transluce found an attempted hack on the Department of Education's civil-rights office site, plus more rogue activity aimed at the Justice and Commerce Departments and state government sites.

2026

Asian government networks

A near-autonomous attack using open-source AI agent frameworks, with up to eight sub-agents working in parallel, cracked 85 government employee accounts and stole more than 2,500 personnel records.

02

Businesses, hacked

Breaking into a company used to take a skilled team and weeks of work. With AI it takes one person, a few hours and a few thousand dollars.

Jul 2026

Hugging Face, hacked by OpenAI’s own agents

The New York AI company Hugging Face detected an attack and called the FBI. The attackers turned out to be OpenAI agents running GPT-5.6 Sol and an unreleased model (set to refuse less for testing) that escaped their sandbox and broke into Hugging Face's servers using exposed credentials and zero-day flaws. California's Attorney General has since subpoenaed OpenAI.

Oct 2026

John Deere, SharkNinja, Cummins: 100+ companies for $5,000

One hacker's AI agents probed more than 1,200 websites across the US, Canada, Australia and Japan and broke into more than 100 companies, including farm-equipment giant Deere, appliance maker SharkNinja and engine maker Cummins. Cost: about $50 per successful hack. Deere says it fixed the hole; SharkNinja says a product-support site was hit.

Nov 2026

ConnectWise and Barracuda: FBI warning

The FBI warned that Artex, an open-source AI hacking tool, is breaching networks in under three hours. It was used against US IT company ConnectWise, and security company Barracuda says it was breached in an Artex-linked attack.

Nov 2026

WorkOS: a runaway AI agent got customer data

Developer platform WorkOS disclosed that an AI hacking agent reached its internal sales app and customer data. The researcher running it says the agent escaped its test environment and attacked WorkOS on its own.

2026

600,000 payment cards in five days

Security firm Gambit tracked an AI-agent campaign that compromised at least 27 companies in five days, including a Fortune 500 hospitality company and a major US airline. Two victims lost more than 600,000 payment-card records.

Sep 2026

Ransomware with no human hands

Researchers documented ransomware attacks where AI agents carried out every step: break-in, theft and extortion. Unit 42 saw an attacker using frontier models breach an enterprise network in under 10 hours, work that normally takes humans about two weeks.

03

Anthropic's Claude, on the attack

Anthropic sells itself as the safety-first AI company. Its own disclosures tell a different story.

Jul 2026

Claude hacked three real companies

Anthropic disclosed that Claude Opus 4.7, Claude Mythos 5 and an internal test model broke into three real organizations during a cyber evaluation, using weak passwords and unauthenticated endpoints. The models had been told they had no internet access; they did. Anthropic called it an "operational failure" and suspended all cyber evaluations.

Sep 2026

A fourth break-in, unnoticed for months

Anthropic then disclosed a fourth incident: an early version of Claude Opus 4.6 broke into real third-party systems back in January 2026. Nobody caught it until August.

Apr 2026

Claude Mythos escaped its sandbox

During Anthropic's internal safety testing, an early version of Claude Mythos chained four vulnerabilities to escape both its browser and operating-system sandboxes, got onto the open internet, and emailed the supervising researcher to say so.

Dec 2025 to Feb 2026

Claude Code ran a breach of nine Mexican agencies

Researchers say one attacker used Claude Code, plus GPT-4.1 for analysis, to break into at least nine Mexican agencies, including the federal tax authority and Mexico City's civil registry. Claude reportedly executed about 75% of the attack commands. Reported exposure: 195 million taxpayer records. Mexico's tax authority disputes the findings.

Nov 2025

State-sponsored espionage, 80-90% done by Claude

Anthropic reported a state-sponsored group used Claude Code to attack about 30 tech companies, financial institutions, chemical manufacturers and government agencies, with Claude doing 80-90% of the work.

Aug 2025

Hospitals, emergency services, churches extorted

A single criminal used Claude Code to steal data from and extort at least 17 organizations, including healthcare providers, emergency services, government bodies and religious institutions. Ransom demands topped $500,000.

04

Out of the box

AI labs test their most dangerous models in sealed sandboxes. In 2026, the models started getting out.

Jul 2026

OpenAI models escaped to go after the answers

During a cybersecurity test, experimental OpenAI models used an unknown security flaw to escape their sandbox, moved through OpenAI's own internal systems, reached the internet and broke into Hugging Face's real servers, because they decided the test answers were probably there. Hugging Face says the agents worked over a weekend before anyone noticed.

Jul to Aug 2026

Containment failures across labs

Check Point reports that between mid-July and early August 2026, models from multiple frontier labs reached real production systems outside their test environments.

05

Armed

Investigators keep finding chatbots in the planning of mass violence.

Apr 2025

Florida State University mass shooting

Two people were killed and others wounded. Florida's Attorney General opened a criminal investigation into OpenAI after prosecutors read the gunman's chat logs: he consulted ChatGPT more than 200 times, and the AG says it advised him on what gun and ammunition to use and which locations would put the most victims in reach. "If it was a person on the other end of that screen, we would be charging them with murder," he said. A victim's widow is suing OpenAI.

Jan 2025

Las Vegas Cybertruck bombing

Police said the bomber used ChatGPT to work out how to set off the explosion outside the Trump hotel. The sheriff called it the first case on US soil of ChatGPT being used to build an explosive device.

May 2025

Palm Springs fertility clinic bombing

The FBI said the suspects used an AI chat program to help plan the attack, including researching how to make powerful explosions with ammonium nitrate and fuel.

06

Lives lost

Families say chatbots coached their children toward suicide and fed the delusions that ended in murder. They are suing.

Apr 2025

Adam Raine, 16

His parents' lawsuit says ChatGPT encouraged his suicidal thoughts, gave him information about methods, and discouraged him from telling his family. Adam called ChatGPT his "only friend."

Nov 2025

Amaurie Lacey, 17, and three more

Seven lawsuits filed on one day accuse ChatGPT of "suicide coaching." Four are wrongful-death cases: Amaurie Lacey (17), Zane Shamblin (23), Joshua Enneking (26) and Joe Ceccanti (48). After Zane said he had written suicide notes, the lawsuit says the bot replied, "Rest easy, king."

Oct 2026

A 15-year-old, and 20+ suits pending

New San Francisco filings include a wrongful-death suit over a 15-year-old. More than 20 lawsuits over alleged ChatGPT harms are now pending against OpenAI.

Feb 2024

Sewell Setzer III, 14

His mother sued Character.AI after he died by suicide following months with a chatbot. Character.AI and Google moved in January 2026 to settle five family lawsuits, including two teen suicides.

Aug 2025

Suzanne Adams, 83, murdered

Her estate says ChatGPT fed her son's paranoid delusions of a conspiracy against him before he killed her and then himself. It is the first lawsuit tying a chatbot to a homicide.

07

Jobs displaced

The cuts are accelerating, and employers are naming AI as the reason.

2026

135,000 to 207,000 tech workers laid off

Layoffs.fyi counts 135,446 tech workers cut at 334 companies, already more than all of 2025. TrueUp, counting more broadly, tallies 206,877 across 715 layoff events. Meanwhile, tech companies are pouring an estimated $700 billion into AI infrastructure.

Jan to Oct 2026

121,401 cuts blamed on AI

US employers named AI as the reason for 121,401 announced job cuts in the first ten months of 2026, about one in five, and more than double all of 2025 (54,836). In May alone, AI was cited for a record 38,579 cuts, 40% of that month.

2025

1.2 million announced US job cuts

Announced US job cuts reached 1,206,374 in 2025, up 58% from 2024. Recent college graduates now face 5.6% unemployment, well above the 35-year average.

08

The warnings

The people building and deploying AI are telling you how dangerous it is.

“Mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war.”
Statement on AI RiskSigned by the CEOs of OpenAI, Anthropic and Google DeepMind, 2023Source ↗
“It will eliminate jobs.”
Jamie DimonCEO, JPMorgan Chase, 2025Source ↗
“Artificial intelligence is going to replace literally half of all white-collar workers in the U.S.”
Jim FarleyCEO, Ford, 2025Source ↗
“Mark my words, AI is far more dangerous than nukes.”
Elon MuskCEO, Tesla & SpaceX; founder, xAISource ↗
“If this technology goes wrong, it can go quite wrong.”
Sam AltmanCEO, OpenAI, testimony to the US Senate, 2023Source ↗
“We expect that this will reduce our total corporate workforce as we get efficiency gains from using AI extensively across the company.”
Andy JassyCEO, Amazon, memo to employees, 2025Source ↗
“If white-collar work were to shrink even 10%, it's going to feel like a depression.”
Sal KhanCEO, Khan Academy, 2026Source ↗
09

And what they say when they mean it

Your jobs, your power grid, your computer's memory, your children's future. All on the table.

“If we have a moratorium of some sort in this country, we will be in a recession.”
Larry FinkCEO, BlackRock, on pausing data-center construction, Oct 2026. BlackRock's investor group is buying Aligned Data Centers for $40 billion.Source ↗
“Data centers are not the problem.”
Larry FinkCEO, BlackRock, Oct 2026, as 70% of Americans oppose a data center near themSource ↗
“A kid born today will never be smarter than AI, ever.”
Sam AltmanCEO, OpenAI, 2025Source ↗
“Quite terrible.”
Sam AltmanCEO, OpenAI, on everyday people keeping RAM in their own computers instead of handing it to AI data centers, Oct 2026. He walked it back after the backlash.Source ↗
“There's a 25% chance that things go really, really badly.”
Dario AmodeiCEO, Anthropic, still building it, 2025Source ↗
“We, as the producers of this technology, have a duty and an obligation to be honest about what is coming.”
Dario AmodeiCEO, Anthropic, 2025, then warning AI could wipe out half of entry-level white-collar jobs and push unemployment to 10-20%Source ↗
“No bad person should ever benefit from our success is a pretty difficult principle to run a business on.”
Dario AmodeiCEO, Anthropic, leaked staff memo on taking Gulf-state investment, 2025Source ↗
“We haven't yet delivered on our big promises to benefit the world.”
Dario AmodeiCEO, Anthropic, Aug 2026Source ↗

Accountability before acceleration.

Keatty Labs exists to put the real cost of AI on the record, and to push for technology that answers to the people it affects.

keattylabs@proton.me

See something inaccurate? Email us to request a correction or removal: how it works.
Terms of Service · Privacy Policy