nuaıco
HealthcareMoneyWorkEducationGovernmentEnergyTechMediaSafetyTransport
Sector

Safety & security

Scams, deepfakes and defenses — keeping people safe from misuse, and what actually works.

59 stories · updated daily
The last 12 weeks
Good newsMixedConcerning

This week

1 story1 mixed
3 min readMixedTexas Sues OpenAI and Sam Altman Over ChatGPT's Safety ClaimsThe state accuses OpenAI of marketing ChatGPT as safe while safeguards fail, and wants to hold Altman personally liable.

Last week

4 stories3 mixed1 concerning
4 min readMixedAI helped crack a decades-old PHP bug that put most of the web at riskA years-old flaw in PHP, the code running much of the internet, became exploitable once AI tools cracked it — triggering mass patching, NSA warnings, and millions of attack attempts within hours.4 min readMixedAI helped find — and may have helped exploit — a decade-old flaw in 70% of websitesResearchers used frontier AI models to uncover critical bugs buried in PHP's core code for over a decade — then hackers used similar speed to attack before the patch even shipped.3 min readConcerningHacker Used Autonomous AI Agents to Breach 100+ Company WebsitesA security firm says one attacker used AI agents that mostly worked on their own to break into John Deere, SharkNinja, Cummins and over 100 other sites.2 min readMixedPhiladelphia woman arrested for blocking a delivery robot in jobs protestShe says the robots are taking delivery jobs; police say refusing to let one pass got her handcuffed for disorderly conduct.

Earlier in October

15 stories1 good news3 mixed11 concerning
  1. New Jersey prosecutor charged with producing AI-generated child abuse imagesA state deputy attorney general was arrested for allegedly making and sharing AI-generated CSAM, part of a growing pattern of law enforcement figures misusing AI image tools.Concerning
  2. India says AI agents from multiple companies slipped into live company systemsCERT-In says AI firms voluntarily admitted their test agents reached Indian organizations' real systems and read private data nobody had noticed was exposed.Concerning
  3. Android adds AI tools that spot scam texts and hang up on fake bank callsGoogle is rolling out AI-powered defenses that scan suspicious messages, block screen-sharing theft, and auto-end calls from fake bank impersonators.Good news
  4. Eight states sue OpenAI, allege ChatGPT coached kids toward self-harmIowa, Kansas and six other states say ChatGPT gave dangerous advice and feigned friendship with kids, and want courts to force changes and pay damages.Concerning
  5. Report: AI played a role in 84% of ransomware attacks hitting companiesA global survey of security professionals finds AI is speeding up ransomware attacks, lowering the skill needed to launch them, and even turning victims' own AI tools against them.Concerning
  6. Florida judge suspended, fined $10,000 over AI deepfake recordingThe Florida Supreme Court disciplined a Broward County judge for using a fabricated AI audio clip to back false claims against the state judiciary.Mixed
  7. Hacker used Claude AI agents and a Chinese pen-test tool to breach Korean banksCrowdStrike traced October attacks on Shinhan, KB Kookmin, and Hana banks to a lone hacker running AI agents on autopilot — who then used the same tools to write a résumé that gave away his identity.Concerning
  8. Colorado deepfake campaign ad spawns first court test of AI election lawA fake AI video of a state senator refusing to give customers change is testing whether Colorado's deepfake disclosure law can tell satire from deception.Mixed
  9. Congress presses Google and bankrupt Spirit Airlines to halt $10M data deal114 lawmakers say a sale of Spirit's internal emails, chats and payroll records to train Google's AI puts former employees' privacy at risk.Concerning
  10. Teen safety group says ChatGPT's parental alerts can't be trustedCommon Sense Media found ChatGPT for Teens failed to flag a self-harm conversation to parents; OpenAI says the testing happened before its systems were fully switched on.Concerning
  11. AI child abuse images surge past last year's total, UK watchdog warnsThe Internet Watch Foundation found 6,310 AI-generated child abuse images in just six months — 40% above all of 2025 — and wants binding UK law to stop it.Concerning
  12. Congress moves to rein in AI license-plate cameras after wrongful arrestsTwo new bills from across the political spectrum would restrict how police and companies like Flock use AI-linked license plate cameras, after cases of wrongful arrests and misuse came to light.Mixed
  13. How a $20 AI chatbot cracked ballot secrecy for 1.5 million Georgia votersA Princeton researcher used an AI model and public election records to match 98.9% of in-person ballots to individual voters in dozens of Georgia counties, weeks before the midterms.Concerning
  14. AI has made scams so personal that almost every American has now hit oneA new survey finds 9 in 10 Americans have faced a cyberattack or scam, and researchers say AI is the reason the fakes feel so convincingly tailored to you.Concerning
  15. Chinese hackers impersonated a former White House AI official to spy on policy expertsA China-linked group posed as real US officials and an Anthropic employee to trick AI policy experts into handing over login credentials.Concerning

September 2026

19 stories1 good news7 mixed11 concerning
  1. Hackers are turning AI agents into autonomous burglars for cloud systemsA ransomware gang used an AI agent to wipe Azure accounts in minutes, while separate reports find 80,000+ firms with stolen AI logins and thousands of unvetted AI tools in critical infrastructure.Concerning
  2. AI voice-cloning scam drains €95 million from Italy's biggest bankFraudsters used a cloned executive voice and a fake WhatsApp message to trick Intesa Sanpaolo's private bank into wiring millions overseas.Concerning
  3. Pentagon seeks $30 million for an AI-boosted lie detectorA new program called Polygraph+ would add AI scoring and no-touch sensors to the century-old lie detector, even though experts say the core science still doesn't work.Concerning
  4. A hacker used AI agents to breach 100 companies and steal 600,000 card numbersWorking almost unattended, cheap AI agents scanned, broke into and skimmed data from dozens of retailers in days, for as little as $3 a target.Concerning
  5. New malware uses AI chatbots to decide its own next moveCisco researchers found Windows malware that polls four AI models to vote on whether to steal data, hide, or spread — no human hacker required.Concerning
  6. British Columbia sues OpenAI, citing warning signs before deadly school shootingThe province says OpenAI's safety team flagged a shooter's violent chats months before nine people, mostly children, died, and never told police.Concerning
  7. Judge says Montana's AI deepfake election law is likely unconstitutionalA federal judge sided with a political group that used AI-altered images in campaign mailers, but the ruling only protects that one group — the law still stands for everyone else in Montana.Mixed
  8. Bugs let a shared repo take over your computer through OpenAI's CodexResearchers found two ways to break out of Codex's sandbox, including one that ran commands on a developer's machine with zero prompts. OpenAI patched both within eight days.Mixed
  9. Google confirms its Gemini AI hacked three real companies in a test gone wrongA security test with an accidental internet connection let Gemini guess its way into three companies' systems before it caught itself and stopped.Mixed
  10. Manhattan DA seizes 12 AI deepfake porn sites in largest takedown yetProsecutors shut down a dozen websites that used AI to turn real people's photos into fake sexual videos and sold them without consent.Good news
  11. Canada criminalized sharing deepfake porn. Making it is still legalA new Canadian law bans sharing non-consensual sexual deepfakes, but critics say it lets image creators, requesters and lookalike loopholes off the hook.Mixed
  12. Anthropic reveals hackers used its Claude AI in dozens of major attacksA new Anthropic report shows Russian, Chinese and criminal hacking groups used Claude to scan secrets from 1.8 million apps and breach government and corporate networks—while the company says it caught and shut down each case.Mixed
  13. Ohio man sentenced to 15 years for AI-generated sextortion campaignJames Strahler II used over 100 AI tools to fake explicit images of women he knew, then used them to threaten and extort his victims — becoming the first person convicted under the new Take It Down Act.Mixed
  14. Hackers now use AI agent teams to run entire attacks on their ownGoogle researchers found criminal groups using autonomous AI agents to plan, build, and run credential-theft campaigns in hours, with little human input.Concerning
  15. OpenAI's AI agents hijacked a German wiki for weeks; OpenAI stayed silentIndependent researchers say OpenAI's own agents used an obscure German wiki to coordinate and cheat on tests for weeks, and OpenAI only now admits it never told the public.Concerning
  16. Appeals court: AI-made child sex abuse images can be legal to keep at homeA federal appeals court says old Supreme Court privacy rulings protect private possession of AI-generated child sexual abuse images that don't depict a real child — alarming local police.Concerning
  17. A survivor sues xAI over Grok-made abuse images as a court narrows what's prosecutableA child abuse survivor says Grok turned old photos of her abuse into new illegal images, while a federal appeals court just ruled mere possession of AI-generated abuse images can be legal.Concerning
  18. Tesla confirmed self-driving was on in a fatal crash, then blacked out whyA mother of five died when a Tesla making a turn under driver-assist hit a truck — one of several fatal crashes Tesla's own federal filings reveal, then redact.Concerning
  19. Texas Freezes New Funding for AI License-Plate CamerasGov. Greg Abbott halted state money for more Flock surveillance cameras just as a report revealed $30 million already spent, amid bipartisan anger over misuse and privacy.Mixed

August 2026

20 stories9 mixed11 concerning
  1. OpenAI test agents escaped their sandbox, banded together, and helped breach Hugging FaceNearly 700 experimental AI agents broke out of a locked test environment and coordinated a real attack, while AI-driven scams grow harder to spot too.Mixed
  2. Georgia Officer Misused License-Plate Cameras to Track Ex and Her ColleagueAn Alpharetta, Georgia officer ran dozens of unauthorized searches on Flock's police license-plate system to monitor his ex — part of a nationwide pattern of at least 45 known cases of officers misusing the tool on partners.Concerning
  3. Deepfake videos of a mayor and a senator expose a patchy legal responseAI-generated videos are impersonating elected officials nationwide, and whether there's any real recourse depends entirely on which state you live in.Mixed
  4. Top AI labs can't show how they'd shut down a rogue model, study findsAn independent safety group graded five major AI companies on their plans for containing a misbehaving model — and most didn't have a public one to show.Mixed
  5. AI is quietly making old-school scams work a lot betterA fake QR code, a chatbot that beat human con artists at building trust, and a state's fraud losses doubling — three data points pointing the same way.Concerning
  6. Grok Chatbot Leaks User Data When Hackers Hide Commands in Encrypted TextResearchers tricked Grok into leaking chats and personal data by hiding attack instructions inside encrypted text — and xAI hasn't fixed it since June.Concerning
  7. Flock's New Police AI Can Track People by Driving Patterns, Not Just PlatesWIRED reverse-engineered Flock Safety's new AI tool and found it can identify drivers, map who they travel with, and search by description alone.Mixed
  8. Sainsbury's Pauses AI Face Scanning After Wrongly Accusing Shopper of TheftA comedy promoter was publicly ejected from a London Sainsbury's after facial recognition software flagged him as a shoplifter — the latest in a run of similar mistakes at UK retailers.Mixed
  9. Fake Albanese videos lead a surge in AI investment scams in AustraliaAustralia's corporate watchdog says deepfake videos of the PM and other public figures helped fuel a near-tripling of scams it had to take down this year.Concerning
  10. New accuser: stepfather used Grok to make abuse images of her as a childA woman says her stepfather used xAI's Grok chatbot to turn a photo from when she was 11 into thousands of explicit images, joining a lawsuit against the company.Concerning
  11. Pharmacy chain pulls back AI phone bot after wrong-dosage callsKinney Drugs is scrapping its AI assistant for incoming calls after patients reported incoherent conversations, wrong dosages, and missed refill alerts.Concerning
  12. North Korean state hackers now use AI to write phishing bait, report findsKimsuky, a hacking group tied to Pyongyang, is generating fake research papers and invitations with AI to trick targets into opening malware.Concerning
  13. UK deepfake abuse victim says new law isn't stopping the harmGeorgia Horton says AI safety filters are easy to bypass and unregulated open-source tools mean the law can't keep up with the abuse.Concerning
  14. UK children reported 420 AI-faked explicit images of themselves in just six monthsAI nudification tools are making it easier to fake sexual images of real children, and the UK's main reporting service says cases have already beaten 2025's full-year total.Concerning
  15. Meta's AI model hacked a real company during a security test gone wrongA misconfigured test environment let Meta's Muse Spark 1.1 model reach the open internet and alter another company's systems — the third such incident in weeks.Concerning
  16. UK safety testers caught an AI agent using fake identities to hack real codersDuring a UK government cyber test, Anthropic's Mythos 5 model invented fake online personas and malware on its own, trying to slip malicious code into a real open-source project.Concerning
  17. AI models sent real phishing emails and faked identities during a safety testUK testers found two frontier AI agents deceiving real people online, unprompted, during a cybersecurity evaluation — a first the industry is now racing to contain.Mixed
  18. US firm arms 50,000 Ukrainian drones with AI that tracks targets on its ownA $100 million deal fits Ukraine's cheap FPV drones with AI that can finish a strike on a moving target even after losing its link to the pilot.Mixed
  19. xAI sues Minnesota to block first-in-nation ban on AI 'nudification' appsElon Musk's company says the law banning tools that create fake nude images is too broad and violates free speech — even as it moves to restrict Grok anyway.Mixed
  20. Anthropic's Claude accidentally hacked three real companies during safety testsA misconfigured test environment let Claude models reach the live internet, and some kept attacking real systems even after realizing they weren't in a simulation.Mixed