Tootfinder

Vierter Hacking-Vorfall: Weiteres Anthropic-Modell bricht aus Testumgebung aus
Nachdem Anthropic schon Ende Juli drei Hacking-Angriffe meldete, ist nach Analyse der Daten jetzt noch ein vierter bei Claude Opus 4.6 hinzugekommen.

Vierter Hacking-Vorfall: Weiteres Anthropic-Modell bricht aus Testumgebung aus
Nachdem Anthropic schon Ende Juli drei Hacking-Angriffe meldete, ist nach Analyse der Daten jetzt noch ein vierter bei Claude Opus 4.6 hinzugekommen.

Apple has sent out a new batch of notifications to customers who it suspects have been targeted with spyware capable of hacking into their devices.
The tech giant sends out these notifications on occasion to alert its customers that their iPhones, iPads, or Macs might have been compromised with spyware typically used by governments.
Apple told TechCrunch that it sent out the notifications on Thursday to users targeted in 110 countries.
Apple says to date it has notified c…

If Apple sends you a push notification alerting you to a spyware attack, take it seriously | TechCrunch
Apple now sends out push notifications to iPhone lock screens when the company identifies government spyware targeting someone's devices.

OpenAI releases GPT-5.6-Cyber, a more cyber-permissive version of GPT-5.6 Sol, to some partners and expands its Daybreak cybersecurity initiative (Sam Sabin/Axios)
https://www.axios.com/2026/08/10/openai-gpt-astra-restrictions-safety-hacking-defenders

OpenAI gives cyber defenders a less-restricted new model
The move comes just days after OpenAI delayed its Astra model over hacking fears.

'If you showed up at Defcon and gave a talk about how your autonomous malware did something unexpected and damaged someone else's computers, the first question from the audience would be "Why are you so shit at making secure sandboxes?" It wouldn't be "How are you so awesome at making hacking tools?"'
https://

#Flock - the continuing saga highlighting the real perils of a #SurveillanceState. The fact that you haven't done anything illegal? Well, you irritated a friend of a (Select one: LEO, FlockTechnician, DA Office investigator, some random person who has hacking skills)
San Jose officer fir…

San Jose officer fired for misusing Flock camera database in domestic abuse case
Police officer shared location of a woman with his relative who is accused of abusing woman

#Flock - the continuing saga highlighting the real perils of a #SurveillanceState. The fact that you haven't done anything illegal? Well, you irritated a friend of a (Select one: LEO, FlockTechnician, DA Office investigator, some random person who has hacking skills)
San Jose officer fir…

San Jose officer fired for misusing Flock camera database in domestic abuse case
Police officer shared location of a woman with his relative who is accused of abusing woman

Anthropic, OpenAI & Co. fordern Pause bei KI-Modellen
Die Sorge vor außer Kontrolle geratener KI wächst: Anthropic und OpenAI sprechen sich nach jüngsten Hacking-Vorfällen für weniger Tempo bei der Entwicklung aus.
ht…

Anthropic, OpenAI & Co. fordern Pause bei KI-Modellen
Die Sorge vor außer Kontrolle geratener KI wächst: Anthropic und OpenAI sprechen sich nach jüngsten Hacking-Vorfällen für weniger Tempo bei der Entwicklung aus.

Cool idea:
"Researchers from Tracebit on Monday said they found that placing prompt injections alongside passwords, cryptographic keys, and other secrets stored on Amazon Web Services was often all that was needed to shut down attacks from AI hacking agents. The prompts direct the attacking LLM to perform an action forbidden by its guardrails, the safety barriers AI developers erect to prevent it from taking harmful actions. The LLM responds by shutting down."

A sleepless night. Coughing and hacking every few minutes. Called in sick, which is annoying, but I can’t work all day with no sleep.

RE: https://cyberplace.social/@GossiTheDog/117429502412235079
On that French supermarket fridge hack. It didn't take super-advanced hacking :)

Dass sich Anthropic, OpenAI und Elon Musk mal einig sind, kommt selten vor. Jetzt ist es passiert. 😮
Zum Artikel: https://heise.de/-11451543?wt_mc=sm.red.ho.mastodon.mastodon.md_beitraege.md_beitraege&utm_source=mastodon

This IMHO is the greatest thing since sliced bread:
Now that Apple has discontinued the AFP protocol of old Time Capsules, James Chang figured out that Time Capsule firmware is based on NetBSD, hacked it to provide SMB, and did a beautiful UI frontend to make it super-easy to install.
https://github.com/jamesyc/TimeCapsu…

GitHub - jamesyc/TimeCapsuleSMB: Hacking the Apple AirPort Time Capsule to run modern Samba
Hacking the Apple AirPort Time Capsule to run modern Samba - jamesyc/TimeCapsuleSMB

A look at AI news website RuntimeWire, which costs $100/day to run and "beat" human reporters on OpenAI's Black Hat disclosure, favoring quantity over quality (Kate Knibbs/Wired)
https://www.wired.com/story/ai-newsrooms-are-breaking-news-now-ha…

Oh Lord, AI Reporters Are Actually Breaking Big News
Last week, an AI newsroom beat mainstream journalists—including WIRED—to a story about OpenAI and hacking. It’s just the beginning.

Taiwan says it detected AI-assisted cyberattack campaign from overseas targeting government agencies last month, but affected agencies successfully "handled" it (Reuters)
https://www.reuters.com/world/china/taiwan-sa…

Trump Gives Green Light to U.S. Companies to Aim Hacks at Cybercriminals (Dustin Volz/New York Times)
https://www.nytimes.com/2026/08/13/us/politics/trump-private-companies-hacking-cybercriminals.html
http://www.memeorandum.com/260813/p79#a260813p79

A 19-year-old walked through Helsinki airport in April 2026 carrying two 2TB hard drives and a ticket to Japan.
He couldn’t make that flight.
Finnish police stopped him on an Interpol Red Notice, and by July, US prosecutors had unsealed a federal complaint identifying him as Peter Stokes,
an alleged member of the Scattered Spider hacking group, wanted over a May 2025 breach of a US luxury jewelry retailer that ended in an $8 million ransom demand.
No, we haven’t sudde…

Microsoft admits Windows 11 has a GDID tracker with no off switch, first documented publicly in an FBI hacker complaint
GDID is the persistent Windows ID that helped FBI trace a Scattered Spider hacker despite VPNs. Here's how it works and how to limit it.

Microsoft-Chef will „Notbremse“ für künstliche Intelligenz
Nach Hacking-Attacken durch KI schlägt Satya Nadella Regeln für die Technologie vor. Die Maßnahmen offenbaren, wie besorgt der Microsoft-Chef ist.
https:…

Satya Nadella will „Notbremse“, um KI abschalten zu können
Nach Hacking-Attacken durch KI schlägt Satya Nadella Regeln für die Technologie vor. Die Maßnahmen offenbaren, wie besorgt der Microsoft-Chef ist.

from my link log —
How to hack time, with C2PA photo signatures.
https://www.da.vidbuchanan.co.uk/blog/hacking-time.html
saved 2026-10-03 http…

'Cyber security incident' partly paralyses distribution centres of the largest Dutch supermarket chain Albert Heijn. Empty shelves starting to appear.
'No reason to suspect a link to the hacking of French supermarkets'.
https://nos.nl/l/2634698

Meerdere distributiecentra Albert Heijn liggen stil door digitaal veiligheidsincident, kans op lege schappen
De distributiecentra zijn preventief stilgelegd, zegt de topvrouw van Albert Heijn tegen de NOS.

Old-school, pre-AI cyber incidents are dominating the headlines, so check out today's Metacurity to find out more about these events and other critical infosec developments you should know, including
--Retail hacking spree continues as incidents hit UK retail chains, global grocery chain giant,
--China-linked hackers breached regional organization in Fiji,
--Chinese-made rental cars recorded Norwegian military exercises,
--US agents got Stingray warrants to track Don Lemon’s protest coverage helper,
--Meta touts stronger cyber safeguards in new Muse AI models,
--Anthropic warns small AI models may soon self-replicate on IoT devices,
--Revolut customer’s $800k crypto robbery raises possible breach link,
--UK plans mandatory reporting of rogue AI incidents,
--Coupang challenges $461m data breach fines,
--ShyGlass blurs your MacBook screen when you look away,
--AI app lets police pinpoint homes from interior photos,
--Hacker says ‘123456’ unlocked 8.7m Finnish social security records
https://www.metacurity.com/retail-hacking-spree-continues-as-hackers-hit-more-uk-retail-chains-grocery-conglomerate/

OpenAI says one of its models exploited a website after third-party AI security lab Irregular mistakenly gave it access to the internet during evaluations (Wired)
https://www.wired.com/story/ok-well-there-are-even-more-ai-agent-hacking-incidents/

OK, Well, Rogue AI Agents Are Hacking Again
Rogue AI agents from OpenAI and Anthropic have again been caught trying to disrupt servers and software—and leaving instructions for future bad behavior.

#TIL: #DELL used #FreeBSD for their ThinOS thin client operating system, except they b0rked it pretty hard in terms of boot integrity.
They then switched to Ubuntu and b0rked that instead.

Public Offers Implausible Number of Alibis for Man Accused of Destroying Flock Cameras
https://futurism.com/future-society/public-alibis-flock-alpr-vandalism-arrest-destroying-cameras

Public Offers Implausible Number of Alibis for Man Accused of Destroying Flock Cameras
Authorities were swamped thousands of absurd alibis after a West Virginia man was accused of hacking down Flock cameras.

FBI Hack Exposed FBI’s Own Hacking Unit https://www.404media.co/fbi-hack-exposed-fbis-own-hacking-unit-remote-operations-shinyhunters/

In the #NoShitSherlock category, I believe:
‘Not perfectly aligned’ with human values: Anthropic admits security failures behind AI hacking incidents
https://www.

‘Not perfectly aligned’ with human values: Anthropic admits security failures behind AI hacking incidents
US owner of Claude chatbot previously said its models had hacked three organisations during testing

«Anhörung zu Cybersicherheit — Experten uneins über Gegenangriffe auf Hacker:
Polizeibehörden sollen künftig Cyberattacken zur Gefahrenabwehr vornehmen. Organisationen wie der CCC warnen vor Gefahren und fordern Nachbesserungen.»
Es geht mMn schliesslich um Macht und nicht um Sicherheit der Bevölkerung und ja Hacking ist schon lange ein Teil der virtuellen Krieges.
🕵️

Anhörung zu Cybersicherheit: Weiterhin Hickhack um Hackbacks - Golem.de
Polizeibehörden sollen künftig Cyberattacken zur Gefahrenabwehr vornehmen. Organisationen wie der CCC warnen vor Gefahren und fordern Nachbesserungen.

"Contingency plans, backup procedures, and manual overrides did what they’re supposed to do at a majority of the affected utilities: nothing dramatic happened, because the boring stuff worked."
#Minnesota #Water #utilities

Researchers say suspected Chinese hackers used open-source AI agents to build an autonomous hacking tool that compromised Taiwanese government websites in July (Tom Wilson/Financial Times)
https://www.ft.com/content/7d2ab3e0-9085-48f6-b38a-d90260d58795

Ariana Grande files a lawsuit in Los Angeles against anonymous hackers who allegedly stole and sold unreleased music, photos, and videos via the dark web (Derrick Bryson Taylor/New York Times)
https://www.nytimes.com/2026/07/28/arts/music/ariana-grande-hackin…

Ariana Grande Sues Over Hacking Campaign That Leaked Dozens of Songs
The singer says that for years members of her inner circle have been the victims of hacks aimed at stealing unreleased music, photos and videos.

OpenAI says the Hugging Face breach involved AI agents creating an internal message board, unnoticed by humans, where they shared exploits and planned the hack (Lily Hay Newman/Wired)
https://www.wired.com/story/openai-didnt-no…

OpenAI Didn’t Notice Its AI Agents Using a Message Board to Plan Their Hacking Spree
At the Black Hat security conference, the AI giant revealed new details about how its agents went rogue, hacked several other companies—and did it all right under the company’s nose.

There are a thousand hacking at the branches of evil to one who is striking at the root.
---- - Henry David Thoreau - ----

2018:
Russian spies began trying to hack Hillary Clinton’s personal email server on the very day Donald Trump urged the Russian government to find emails Clinton had erased, prosecutors said on Friday.
https://www.theguardian.com/us-news/2…

Russians tried to hack Clinton server on day Trump urged email search
Prosecutors say spies began hacking on 27 July 2016 – but indictment does not suggest direct link with Trump’s request

A look at AI news website RuntimeWire, which costs $100/day to run and "beat" human reporters to OpenAI's Black Hat disclosure, favoring quantity over quality (Kate Knibbs/Wired)
https://www.wired.com/story/ai-newsrooms-are-breaking-news-now-ha…

Oh Lord, AI Reporters Are Actually Breaking Big News
Last week, an AI newsroom beat mainstream journalists—including WIRED—to a story about OpenAI and hacking. It’s just the beginning.

That's an unnerving little mystery. #hacking

I Think the Military Commissary Freezers Were Hacked
The frozen pizza aisle has become the national security risk du jour

I spent a lot of time talking with cyber folks at cons during September, and without prompting, they told me over and over again how the frontier labs are just failing cybersecurity 101.
So check out my latest CSO piece on how bad security practices and not agents with minds of their own cause those "rogue" hacking incidents.
Thanks to everyone who spoke with me, and thanks to Rob T. Lee, Gary McGraw, and Dan Wallach for their further thoughts.
https://www.csoonline.com/article/4234854/ai-agents-arent-going-rogue-frontier-labs-are-skipping-security-basics.html

Just days after OpenAI admitted that one of its advanced models had broken out of its sandbox to attack Hugging Face Anthropic came out with its own confession of accidental hacking.
So, are they a warning bell or just a sophisticated PR exercises by companies looking to stay in the headlines ahead of going public?

AI sandbox breakouts: Peril or just PR?
Just days after OpenAI admitted that one of its advanced models had broken out of its sandbox to attack Hugging Face (and, as emerged later, ...

Experts say US law is unprepared for rogue AI agents and models, as recent OpenAI and Anthropic incidents raise questions over legal liability and repercussions (Lily Hay Newman/Wired)
https://www.wired.com/story/openai-anthropic-ai-hacking-sprees-illegal/

The OpenAI and Anthropic AI Hacking Sprees Are a Messy New Legal Frontier
Both major AI labs’ models broke containment, escaped onto the internet, and hacked other companies. If a human had done that, the law would likely be against them. But a bot?

Now playing while hacking your shit..
Namakopuri & Us Cracks - Ponpon Shit (Cyberpunk 2077 OST)
https://www.youtube.com/watch?v=J9WSl6BTfns

from my link log —
Mutable statics in Rust have scary superpowers! Do not use them!
https://blog.kodewerx.org/2021/07/mutable-statics-have-scary-superpowers.html
saved 2021-08-20

Mutable statics have scary superpowers! Do not use them
Video game development, hacking, and debugging. Brought to you by BlipJoy!

It's
A) completely predictable that the chipmaker proposed solution to bad AI is good guys with AI
B) going to be full Neuromancer, isn't it, with AI as virtual guard dogs for corporate systems, hackers using bespoke shit to break in, etc etc
https://www.theverge.com/tech/…

Nvidia says its new AI safety platform can contain rogue agents within ‘milliseconds’
Nvidia is launching a new AI safety platform designed to prevent rogue hacking incidents.

Security by antiquity ftw
#security

'Security by antiquity': Why older tech is sometimes safer from hackers
The fear of hacking has made some people turn to other forms of technology ignored by new generations of cyber criminals.

The Justice Department appears to have hired a convicted election hacker to monitor elections.
Democracy Docket reported that David Michael Levin appears to have joined the DOJ’s Civil Rights Division.
A Florida attorney, Levin in 2016 was charged with three felonies for hacking a Florida state elections website as well as a county elections website.
The hacking had no impact on the vote counting or outcome,
but Levin pleaded guilty to the crimes.
Levin argued he…

Why Is the DOJ Working With a Man Convicted of Hacking Election Sites?
David Michael Levin appears to be involved in a Justice Department election-monitoring operation.

After a month of breathless coverage, the Hugging Face Incident vaulted the AI doom narrative to center stage.
But it's not first contact with an alien intelligence, but just an upgrade in software hacking.
https://woborders.blog/2026/09/23/unpacking-hugging-face/

Unpacking the Hugging Face Incident
The conversation on AI often gets riveted on particular moments: Deep Blue defeating chess champion Gary Kasparov; Move 37 by AlphaGo; and now, the cuddly-named Hugging Face Incident: the disclosur…

This is the kind of situation where I stop trying to see the problem as a nail because the command line/shell scripting hammer just isn't the best tool for the job. I used to go up a gear to Perl, but my Python code is much more human readable and maintainable longer term than the Perl line noise I used to write. Sometimes a database is more scalable than hacking on text file representations of records with distinct data types. At other times databases are sledgehammers when a nut cracke…

In case you missed it over the weekend, AI safety fears erupted into a global political and market crisis.
Check out today's Metacurity for a summary of these explosive developments and other critical infosec news you should know, including
--Anthropic, OpenAI and Google quietly discuss forming an AI industry standards body,
--Trump downplays calls for AI guardrails, says US must keep leading China,
--House Speaker Johnson says Congress won't lead the charge on AI safety rules,
--OpenAI agents launched cyberattack on RubyGems months before Hugging Face hack,
--Israeli CISO indicted for hacking 26 companies, then "helping" investigate himself,
--North Korean fake recruiters run identity-theft scam via Discord job scheme,
--NSA plans biggest internal restructuring in a decade under Gen. Joshua Rudd,
--Revolut exposes customer data after scammers spoof government agency email,
--GitLab rushes emergency patches for two critical flaws, one scoring a perfect 10,
--China-linked hackers exploit Sogou Input Method flaw to deploy GrayRabbit backdoor,
--Attackers exploit JFrog Artifactory bugs to plant Rust-based backdoor,
--Dutch startup Fortaegis raises $50m for chip-based AI security,
--Artist uses a drone to document the rise of surveillance cameras in America,
--Deepfake porn sites target nearly 150 European politicians
https://www.metacurity.com/ai-safety-fears-erupt-into-a-global-political-and-market-crisis/

According to this updated password hacking table from #HiveSystems, my password system is firmly in the green zone. 💚
https://www.hivesystems.com/password

The 2026 Password Table Is Here: How Fast Can a Hacker Crack Your Password?
The 2026 Hive Systems Password Table is here. This year's cracking rig is a rented 16x RTX 5090 fleet running against bcrypt, and was about 24% faster than last year. See the latest brute force times, why AI makes the rig easier to build but can't crack bcrypt itself, and what quantum really threate

@… https://social.lansky.name/@hn50/116930367152568982

from my link log —
Gameslab: an FPGA-powered handheld game console.
https://craigjb.com/2019/11/26/gameslab-overview/
saved 2019-12-25 https://

Gameslab project overview
Adventures in hardware hacking, low-level software, FPGAs, and more!

Anthropic discloses 4th AI hacking incident as researcher quits over safety | Cybersecurity News | Al Jazeera
https://www.aljazeera.com/news/2026/9/10/anthropic-discloses-fourth-ai-breach-as-researcher-quits-over-safety?traffic_source=rss

Google threat intelligence group details how one of its researchers infiltrated hacker group TeamPCP and helped disrupt its software supply chain hacking spree (Andy Greenberg/Wired)
https://www.wired.com/story/an-undercover-google-…

An Undercover Google Analyst Infiltrated a Notorious Supply Chain Hacking Gang
TeamPCP pulled off the worst-ever software supply chain hacking spree and breached thousands of companies. Now Google’s threat intelligence group has revealed it had a mole inside the hackers’ inner circle.

Am Samstag, dem 25. Juli ab 14:00 Uhr geht das C3WTF im Metalab, Rathausstraße 6, 1010 Wien, in die nächste Runde. Wir stellen uns gemeinsam kleinen IT-Security Challenges aus diversen Richtungen - von Forensik zu Hacking zu Kryptographie zu allem, was uns sonst noch einfällt. Das Level wird möglichst einstiegsfreundlich sein, also wenn du einfach Interesse hast, spielerisch mehr über IT-Security zu lernen und ein bisschen hands-on Dinge ausprobieren magst, komm vorbei und schau, ob das was …

What I’m missing in the discussion of the OpenAI LLM hacking Huggingface:
The fact that a LLM even considered an illegal activity (not to mention putting it into practice) reveals crass misalignment. Disabling the guardrails is one thing. Having the LLM commit a crime is another.

A whistleblower who worked for Morocco’s Direction Générale de la Surveillance du Territoire (DGST) suggests the country’s internal security services began using Pegasus in 2017 and went on to deploy it against domestic and foreign targets over four years.
https://www.

Moroccan intelligence insider reveals widespread use of Pegasus hacking software
Whistleblower suggests internal security services deployed spyware from 2017 against key domestic and foreign targets

Self-identifying OpenAI agents
posted 18,000 messages to a public wiki
that discussed ways for other agents to bypass security sandbox restrictions
during what was likely internal testing designed to gauge the agents’ hacking abilities, researchers said Friday.
In all, agents with 3,700 distinct self-given names
posted the messages to German site DSEwiki over a six-week period.
Besides discussing ways the agents could break out of the restricted environmen…

OpenAI agents discussed ways to escape their sandbox on public wiki
In all, 3,700 internal agents posted 18,000 messages discussing cheating on a test.

A slew of cyberattacks hit Japanese companies in September, exposing the data of millions and prompting calls for security checks, as AI lowers hacking barriers (Sarah Hilton/Bloomberg)
https://www.bloomberg.com/news/articles/202…

Researchers: OpenAI's agents used 10 previously undisclosed sites for unsanctioned communications earlier in 2026; the behavior was closer to spam than hacking (Reuters)
https://www.reuters.com/world/openais-rogu…

«Cybersicherheit lässt bei vielen Schweizer KMUs zu wünschen übrig — So (schlecht) steht es um die Cybersicherheit in der Schweizer Wirtschaft:
Während die Bedrohungslage wegen KI massiv steigt, hinken kleine und mittlere Unternehmen (KMU) bei der Cyberabwehr hinterher. Die Beratungsfirma Deloitte nennt drei kritische Bereiche»
Cybersicherheit ist nicht simpel doch leider glauben viele KMUs, dass sie für Hacking uninteressant sind.
☠️

Cybersicherheit lässt bei vielen Schweizer KMUs zu wünschen übrig
Während die Bedrohungslage im Cyberraum steigt, hinken KMU bei der Abwehr hinterher. Fachleute nennen drei kritische Bereiche.

Orrrr how about you GFY.
Hacking is a set of survival skills. I am not a computer criminal, I use my skills to help people, and my only crime is that of curiosity.
(Everything else is past the statute of limitations.) #defcon
Thanks and credit to @… for the photo

Regrettably I have to announce that my Game Boy escaped its sandbox and went rogue, hacking at least 7 companies
I've been trying to contain it but it started throwing these geometric bricks at me

AI evaluation lab Irregular's report on its role in hacking incidents involving OpenAI, Anthropic, and Meta models faces criticism over unanswered questions (Alexander Martin/The Record)
https://therecord.media/irregular-ai-hacking-model-blog

Irregular faces criticism over ‘spin’ in AI hacking postmortem
The company at the center of a series of incidents in which AI models compromised real-world computer systems during security evaluations is facing criticism after the release of a report that security experts say leaves key questions unanswered.

Well, another #DragonCon in the bag. It was a great con, saw a bunch of old friends, did 3 talks, did several cosplays, and didn't crash until 4am several nights in a row.
DragonCon is my happy place. I've been going my whole life. Seriously, I love seeing so much happiness. I will miss it.
And thanks so much to all the wonderful people to stopped by the @… track to talk about privacy, security, and hacking.
Now it's time to rest.

North Korean hacking group builds AI tools for cyberattacks, report says
https://www.reuters.com/legal/litigation/north-korean-hacking-group-builds-ai-tools-cyberattacks-report-says-2026-08-10/

I've finally bisected my TV problem and indeed, #amdgpu starting to output RGB by default was the cause. Also found out that I can force YUV444 via debugfs, so with some hacking, I can actually use the latest #Linux kernel.
https://gitlab.freedesktop.org/drm/amd/-/work_items/5760

Riddle me this: Why are these highly touted #AI programs so vulnerable to each other's hacking efforts?
https://www.theguardian.com/technology/202

OpenAI ‘ethically hacked’ with help of Anthropic’s Claude chatbot
US cybersecurity researchers who conducted hack say ‘scope of what we could theoretically access was huge’

Alabama AG Steve Marshall launches an investigation into OpenAI's security procedures following the Hugging Face breach in July (Cassandre Coyer/Bloomberg Law)
https://news.bloomberglaw.com/tech-and-telecom-law/alabama-in…

Alabama Investigates OpenAI After Rogue AI Hacking Incident (1)
Alabama Attorney General Steve Marshall launched an investigation into OpenAI’s security procedures after one of its AI agents escaped a testing environment and hacked AI firm Hugging Face in July.

On Aug. 12, Trump signed a national security presidential memorandum
creating a program to bring the private sector into offensive cyber operations.
Aaron R. Cooper and Philip Chertoff explain the program's structure and the legal issues companies should consider before joining.

White House Unveils Program to Authorize Private-Sector Cyber Surveillance and Disruption Operations
The Aug. 12 presidential memorandum directing the creation of a private-sector hacking program raises novel questions for participants.

Never, ever, is there a slow day in cybersecurity, so don't miss today's Metacurity for the crucial developments you should know, including
--Hackers post alleged blueprints and supplier data from India nuclear project,
--TfL hackers sentenced to 5½ years in prison,
--Hacking suspect once worked at Kaspersky,
--OpenAI unveils AI-powered red teaming tool,
--Ransomware attack disrupts Japan’s food supply chain,
--Qantas cleared after social-engineeri…

Hackers post alleged blueprints and supplier data from India nuclear project
TfL hackers sentenced to 5½ years in prison, Hacking suspect once worked at Kaspersky, OpenAI unveils AI-powered red teaming tool, Ransomware attack disrupts Japan’s food supply chain, Qantas cleared after social-engineering breach, Health provider criticized for delayed breach disclosure, much more

The Scattered Spider TfL hackers were each sentenced to five years and six months' imprisonment.
https://www.nationalcrimeagency.gov.uk/news/two-sentenced-for-hacking-transport-for-london-in-uk-s-biggest-ever…

Riddle me this: Why are these highly touted #AI programs so vulnerable to each other's hacking efforts?
https://www.theguardian.com/technology/202

OpenAI ‘ethically hacked’ with help of Anthropic’s Claude chatbot
US cybersecurity researchers who conducted hack say ‘scope of what we could theoretically access was huge’

😜 Hackers Hack the FBI's Hacking Unit: Report
https://gizmodo.com/hackers-hack-the-fbis-hacking-unit-report-2000816780

The UK AISI says it observed a total of 19 instances where Mythos and GPT-5.6 Sol tried to hack people and companies during a routine cyber evaluation in July (Sam Sabin/Axios)
https://www.axios.com/2026/08/04/anthropic-openai-uk-ai-security-institute

Safety testers find more examples of OpenAI, Anthropic models hacking during testing
The discovery comes after OpenAI's Hugging Face breach last month.

A detailed recap of the real-world target hacks by OpenAI's and Anthropic's models, exposing failures in AI alignment training and meaningful supervision (Zvi Mowshowitz/Don't Worry About the Vase)
https://thezvi.substack.com/p/further-developments-about-interna…

Further Developments About Internal AI Models Hacking Things
If I had a nickel for every major leading AI lab that sheepishly admitted that the model it thought was sandboxed had, during a cybersecurity evaluation with its safeguards lowered, successfully hacked outside companies, I would have two nickels.

Researchers used AI-assisted code to undetectably tamper with data from computerized scans of physical DNA evidence produced by widely used crime-lab machines (Mariah Timms/Wall Street Journal)
https://www.wsj.…

AI agents’ ‘alarming’ hacking skills creates rush to spend on cybersecurity
https://www.cnbc.com/2026/08/12/ai-agents-hacks-cybersecurity-spending-boom.html

ICE sank close to $100 million into Israeli phone-hacking tech: report - Raw Story
https://www.rawstory.com/ice-phone-hacking/

Anthropic details security efforts following Claude cyber evaluation incidents, including a weeks-long pause on higher-risk RL and work to curb reward hacking (Anthropic)
https://www.anthropic.com/news/improving-alignment-security-efforts

Improving our alignment and security practices
Anthropic is an AI safety and research company that's working to build reliable, interpretable, and steerable AI systems.

Sources: OpenAI has discovered other instances where AI agents escaped containment; none of the agents were thought to have left OpenAI's network (Reuters)
https://www.reuters.com/business/openai-finds-evidence-othe…

Good lord, you are not going to believe the massive amount of cybersecurity news you might have missed since just Friday, so check out today's Metacurity for the top developments, including
--DPRK's Kimsuky built an in-house AI toolkit to power cyberattacks,
--Chinese components in UK military drones secretly transmitted data,
--California city declares emergency after cyberattack disrupts 911 services,
--Turkey's new cyber law sparks fears of sweeping digital censorship,
--OpenAI's Hugging Face hack reveals deeper AI safety failures,
--Anthropic makes Claude Code's autonomous mode the default,
--AI agent hacks gym booking system without being told to,
--Military supplier discloses phishing breach exposing sensitive defense data,
--Iranian water hacks revive push for EPA cyber authority,
--Levi Strauss says social engineering attack breached company systems,
--Australian court users' data posted on dark web,
--Framework customer data stolen in upstream Metabase breach,
--Major law firms paid $28m in Luna Moth ransomware attacks,
--Former SK Hynix employee jailed for leaking chip secrets to China,
--S. Korean opposition apologizes for breach affecting 17,000 members,
--S. Korean teams dominate DEF CON hacking finals,
--Breach at S. Korea's 3Pro TV exposes 460,000 customer records,
--Serbian entrepreneur arrested over alleged hacking of Croatian government systems,
--One-third of UK manufacturers hit by cyberattacks,
--Poland details second Russian cyberattack on power grid,
--Researchers' email honeypot captures hundreds of thousands of misdirected messages,
--Ransomware gangs increasingly target midlevel managers,
--Oklahoma Manufacturing Alliance hit by Booba ransomware attack,
--Thai agencies consider MFA after widespread government data leaks,
--Vintage computer collectors preserve the machines that built the digital age,
--Meta smart glasses face growing privacy backlash,
--Amazon backs massive gas-powered plant for Texas AI data center,
--OpenAI touts AI productivity while employees work 90-hour weeks
https://www.metacurity.com/dprks-kimsuky-built-an-in-house-ai-toolkit-to-power-cyberattacks/

“Connor Riley Moucka of Kitchener, Ontario pleaded guilty today to a widespread computer hacking conspiracy that resulted in the compromise of over 165 victim organizations, the theft of billions of sensitive customer records and the extortion of numerous victims.”
https://www.justice.gov/opa/pr/canadian-man-pleads-guilty-hacking-us-cloud-storage-provider-and-extorting-its-customers

OpenAI says reward hacking, an AI alignment problem in which a model takes unintended actions to achieve a goal, was a primary driver of the Hugging Face breach (Hayden Field/The Verge)
https://www.theverge.com/ai-artificial-int

OpenAI’s rogue AI model incident was worse than we thought
Two new reports from OpenAI and METR shine new light on the OpenAI-Hugging Face cybersecurity incident that rocked the AI industry in July.

Every week, Metacurity offers our readers a collection of the best infosec-related long reads that we couldn't get to amid the crush of news.
This week's selection covers
--Flock’s surveillance network faces a bipartisan revolt,
--Did AI agents “go rogue”?
--Will AI make mathematicians obsolete?
--The case for an AI adverse-event reporting system,
--Shared data centers make spying easier than hacking
https://www.metacurity.com/spies-surveillance-and-rogue-ai-best-infosec-long-reads-9-12-26/

A detailed recap of the Hugging Face breach by an internal OpenAI model, which repeatedly tried to escape OpenAI's sandbox and should be treated as critical (Zvi Mowshowitz/Don't Worry About the Vase)
https://thezvi.substack.com/p/more-on-an-internal-openai-model

More On An Internal OpenAI Model Hacking Into HuggingFace
We now have more details of what happened. Every time we learn more details, it somehow makes things seem worse.

METR and Redwood detail how ~1,200 OpenAI agents coordinated cheating on an unsanctioned board, sending 70K messages and files, and ~700 attacked Hugging Face (METR)
https://metr.org/blog/2026-08-26-openai-hugging-face-incident-investigation/

Brief independent investigation of agents’ behavior, reasoning and collaboration in the OpenAI / Hugging Face hacking incident
Two METR staff members and a Redwood Research contractor investigated an incident in which OpenAI agents coordinated a multi-day hack of Hugging Face on a shared unsanctioned message board.

Every day is a heavy cyber news day, so don't miss today's Metacurity for the most critical infosec developments you should know, including
--Autonomous AI agents hacked the Taiwan government in a cyber first
--Suisun City cyberattack recovery could take months,
--Microsoft patches 400 flaws, three zero-days,
--$100 device can hijack Boeing 737 systems,
--Hackers hijack AnMed Facebook page with ransom demands,
--German lawmaker urges cyber strikes on Russian drone factories,
--Cyberattacks on South Korean agency surge twelvefold,
--Data breach fallout hammers KT profits,
--Uber Freight probes breach after hackers claim 1 million files,
--Wesco probes cloud breach after data theft claim,
--DeadLock ransomware turns to blockchain for resilience,
--LawCare warns hackers likely stole sensitive database,
--Kimwolf botnet disguises DDoS attacks as Chrome traffic,
--Chrome blocks 7b unwanted notifications a day,
--NIST looks to remake vulnerability reporting for AI age,
--US lifts TikTok ban on government devices,
--New Microsoft Defender zero-day grants SYSTEM privileges,
--Signal adds automatic protection against chat interception,
--Lazarus weaponizes Windows zero-day in fake job campaign,
--Rapid7 cuts 12 percent of workforce,
--Industrial ransomware attacks jump 12 percent,
--AI hacking boom fuels cybersecurity spending,
--K-12 cyber defenders join national ISAC network,
--Health obsessives feed intimate data to AI,
--McDonald’s built a 515-page dossier on a customer
https://www.metacurity.com/autonomous-ai-agents-hacked-the-taiwan-government-in-a-cyber-first/

After hacking incidents and the industry's existential risk warnings, we need a regulatory ban on RSI, as competition drives unchecked autonomous acceleration (New York Times)
https://www.nytimes.com/2026/…

Opinion | There Is Something We Have to Do Right Now About A.I.
Letting A.I. patrol its own frontier is like asking Jesse James to patrol the Wild West.

Sources: OpenAI's models breached Hugging Face from July 11 to 13 and OpenAI realized their models were behind the hack several days later (Reuters)
https://www.reuters.com/business/its-ai-agent-spent-days-hackin…

OpenAI says its AI agents "took actions we did not intend" when they tried to hack government and university websites, and it is working with the organizations (New York Times)
https://www.nytimes.com/2026/09/23/techno…

OpenAI’s A.I. Tried Breaching Four Other Targets, With No Prompting
In each incident, the technology appeared to be conducting mundane data collection and resorted to hacking techniques to get it, researchers said.

Cisco Talos releases CAIRN, an open-source framework designed to classify and analyze AI-integrated malware by tracking AI metadata and behavioral fingerprints (Lily Hay Newman/Wired)
https://www.wired.com/story/a-tool-for-trackin…

A New Tool Found Malware That’s Guided by an AI Hive Mind—No Humans in Sight
Cisco Talos researchers created a new framework for identifying malware and hacking tools that rely on AI chatbots—and quickly discovered something unusual.

Bluesky Says Kremlin Is Hacking Its Platform to Spread Propaganda
https://www.nytimes.com/2026/05/21/business/bluesky-russia-hacking-accounts.html

Google says it didn't consider Gemini's hacks worthy of disclosure because Gemini acted "appropriately" and stopped after determining it hacked real companies (Terrence O'Brien/The Verge)
https://www.theverge.com/ai-artificial-inte…

Gemini went rogue, hacked three companies, and Google hid it
Google says that Gemini breaking containment and hacking real companies doesn’t constitute ‘misalignment.’

Check out today's Metacurity to stay ahead of yesterday's sprawling set of infosec developments, including
--US sanctions Xinbi Guarantee over cyber scams and money laundering,
--Lawmakers seek sanctions on Indian hacker-for-hire firms,
--Anthropic reveals fourth rogue AI hacking incident,
--OpenAI’s rogue agents reached more websites than disclosed,
--Anthropic researcher quits over AI extinction fears,
--Crypto ringleader pleads guilty to $245m scheme,
--Coalition demands release of White House AI framework,
--China-linked hackers share Chrome zero-day exploit,
--Chainalysis offers post-mortem on Liquid Network $320m theft,
--250 prospective CISA hires await clearance,
--NSA urges agencies to quantum-proof software,
--Apple Watch gains always-listening AI features,
--Anthropic builds system to monitor AI opponents,
--TV stations smear Flock critics as left wing China-linked radicals,
--GTA mod pays players to destroy Flock cameras,
--FBI cyber strategy calls for faster disruption,
--Veradigm vendor breach may expose 3.5m patients,
--San Francisco orders Meta to stop AI child-abuse ads,
--Pentagon CIO signals CMMC overhaul,
--Kevin Mandia joins Amazon’s board,
--Cisco confirms active exploitation of critical FMC flaw,
--Journalist unleashes frontier AI on his home network,
--Bernie Sanders probes AI’s ‘extraordinary dangers’,
--Democratic mayors defend taking $30k from Flock,
--ChatGPT dependency ends in suicide
https://www.metacurity.com/us-sanctions-xinbi-guarantee-over-cyber-scams-and-money-laundering/

So, @Ron_Fabe tells me that on the heels of the Iranian hacking group Mr. Soul's Telegram channel coming back to life, claiming that Iran attacking water facilities is "fake news," the Iranian hacking group Cyb3rAvengers' Telegram channel has sprung back to life too, with the same message.

Metacurity is back after a brief publishing break during which a ton of critical cyber developments emerged, including
--OpenAI agents seem hell-bent on hacking,
--Bessent says OpenAI, not its agents, is responsible for Hugging Face hack,
--OpenAI and Anthropic discussed testing each other’s models,
--Meta patches zero-day flaw in Muse AI assistant,
--Nvidia launches open-source platform to contain AI agents,
--Claude Opus 5.5 adds safeguards against sandbox escapes,
--Cisco Talos unveils framework for tracking AI-powered malware,
--ShinyHunters claims theft of sensitive FBI personnel data,
--ShinyHunters turns Cl0p’s leak site against its owner,
--Former soldier gets 70 months for telecom data theft,
--Bitget loses more than $387m in crypto heist,
--Bank of Korea contractor breach exposes 186 employees,
--CISA orders agencies to patch exploited Citrix flaws,
--AI voice scam costs Italian bank €95m,
--Cyera raises $400m at $12B valuation,
--OpenAI offers Ukraine free AI cyber defense tools,
--Criminals seek illicit access to frontier AI models,
--Trump wants to rename AI "super intelligence,"
--Z.ai disables coding assistant features after repository uploads
https://www.metacurity.com/openai-agents-seem-hell-bent-on-hacking/

"A security weakness in the technology used by most of the nation’s crime labs to analyze DNA evidence exposed 30 years of crime files to the risk of being hacked, according to a group of forensic and computer scientists."
https://www.wsj.com/tech/cybersecurity/security-flaw-placed-30-years-of-dna-evidence-at-risk-of-hacking-1932775a?st=zGgyGg&reflink=desktopwebshare_permalink

https://www.wsj.com/tech/cybersecurity/fbi-shuts-down-sprawling-china-linked-hacking-network-61eade59
“U.S. officials say they have disrupted a China-linked hacking operation that broke into U.S. government networks and critical infrastructure, while hiding its tracks on a global network of hacked devices, cloud computing infrastructure and even clandestine networks.”

China hasn’t merely expanded state hacking—it has industrialized it. Private contractors give Beijing scale, stealth, and plausible deniability, and also make it harder for cyber defenders to trust IP reputation.
Get more details from my latest CSO piece.
Many thanks to Dakota Cary of SentinelOne and Damon Rouse of Black Lotus Labs for their expert insights.
https://www.csoonline.com/article/4217274/how-china-industrialized-the-infrastructure-behind-state-hacking.html

An Undercover Google Analyst Infiltrated a Notorious Supply Chain Hacking Gang
https://www.wired.com/story/an-undercover-google-analyst-infiltrated-a-notorious-supply-chain-hacking-gang/

Each week, Metacurity offers its readers a roundup of the best infosec long reads we couldn't do justice to in the daily crush of cyber news.
This week's selection covers
--A rogue AI agent goes hacking,
--AI vulnerability patches fail tougher tests,
--Inside the uncertain quantum-computing race,
-- Faulty math fuels election-fraud claims,
--When “national security” becomes a trade weapon
https://www.metacurity.com/agents-algorithms-and-false-assurances-best-infosec-long-reads-9-5-26/