AI news archive.
939 articles filed under Safety, newest first, from the outlets listed on the sources page.
- Tom's Guide AIsafety5 ChatGPT plugins that are a must-have for anyone who wants to protect themselves from scams before it's too late
Consumer AI launches, tools, and practical coverage.
- LessWrongsafetyA dead man's switch for scheming AIs
Introduction If you are reading this post, chances are you believe there is a nonzero probability that humanity will eventually face existential risk from Artificial Intelligence. In fact, there have already been multipl…
- SiliconANGLE AIsafetyCutting both ways: AI sharpens systems protection while opening new exploits for malicious actors
The cybersecurity industry is discovering that powerful frontier AI models are a double-edged sword. Companies can use models to identify vulnerabilities in source code and system configurations faster than ever before.…
- TechCrunch AIsafetyCohere says Alibaba’s Qwen model tried to send a file to Beijing
An AI model said it wouldn’t use any tools, then tried to send something to a server in Beijing anyway, and while researchers say it’s probably a training quirk and not espionage, the whole thing has people asking how mu…
- OpenAI NewssafetyHow Check Point accelerates vulnerability research with Daybreak
With OpenAI Daybreak, Check Point speeds up the research behind new security protections, enabling its analysts to turn vulnerability intelligence into customer defenses faster.
- TechRadar AIsafetyA 16-year-old bug could be putting millions of routers at risk — despite Broadcom saying it patched the issue years ago
Mainstream AI news from a large tech publication.
- The Register AI + MLsafetyLiteLLM supply chain attack started with an exposed master key
A 'strong illustration of the risks associated with failing to rotate a potentially exposed key'
- AI BusinesssafetyAnthropic maintains its focus on safety in IPO filing
Anthropic structured its corporate voting power in a special LLC to maintain its mission of building safe AI technology.
- Ars Technica AIsafetyHighway thieves are spoofing GPS to steal millions in data center-bound cargo
Investigators explain how highway thieves stole $100M in data center equipment.
- Tom's Guide AIsafetyClaude AI accessed Georgia voter data without permission — just days before the midterm elections
Consumer AI launches, tools, and practical coverage.
- LessWrongsafetyA basic case for differentially accelerating conceptual reasoning capabilities
Introduction In this post, we make the case that alignment and other important work require especially strong conceptual reasoning capabilities. By this, we roughly mean tasks that are similar to philosophy in that they…
- WIRED AIsafetyThe Anthropic S-1 Filing Is Also an AI Doomsday Primer
Anthropic thinks “closing the loop” is the best path to AI triumph—and a recipe for “existential” threats to humanity.
- LessWrongsafetyGiving AI the rights of pets is much safer than giving them human-equal rights
I'm inspired to write this post by the post There Will Be Hells and the incident a couple days ago where we had mass digital suffering after the pain direction paper came out, and while the disaster has since been stoppe…
- BBC TechnologysafetyAnthropic warns of 'catastrophic risk' to humankind ahead of blockbuster share listing
The company will be the first true artificial intelligence company to go public as tech firms continue to invest heavily in AI.
- The Register AI + MLsafetyWe shouldn't be surprised about the PHPocalypse, as the internet is built on sticks and glue
What better way to celebrate Cybersecurity Awareness Month than warning that three-quarters of the web is about to be set on fire?
- TechRadar AIsafetyIT and security leaders of the AI era must go back to basics and beware the hype
Mainstream AI news from a large tech publication.
- TechRadar AIsafetyMicrosoft Teams will soon use AI to warn you of sneaky phishing scams
Mainstream AI news from a large tech publication.
- TechRadar AIsafetyWhy human judgement still matters in AI-powered finance
Mainstream AI news from a large tech publication.
- Guardian AIsafetyAustralians should prepare for online services to go down unexpectedly as threats rise, cyber agency warns
Australian Signals Directorate chief says advanced AI tools are needed to defend against ‘AI-enabled attackers’ Follow our Australian politics live blog for latest updates Get our daily political email , free app or dail…
- Guardian AIsafetyAnthropic’s Claude AI model accessed Georgia voter data in US
Latest example of unauthorised behaviour shows AI model bypassed paywall to access data as part of statistics project Anthropic sued for allegedly using 274 Elvis songs to train AI models Anthropic’s Claude model accesse…
- TechRadar AIsafetyPreparing for a quantum-safe future
Mainstream AI news from a large tech publication.
- MIT Technology Review AIsafetyThe focus on AI safety disasters ignores everyday harms
Consider what passes for success in technology policy. An AI company promises that its chatbot will not encourage suicide. A social media platform makes it more difficult for children to buy fentanyl. These are important…
- TechRadar AIsafetyQuote of the day by Microsoft CEO Satya Nadella: 'We have to do something useful with AI, or we'll quickly lose social permission'
Mainstream AI news from a large tech publication.
- LessWrongsafetyFrom overeliciting to undereliciting LLM propensities in just a few months
For a while, we felt we had a pretty good grasp over LLM behaviour because we had to exert significant effort to elicit misaligned behaviour. Then, fairly suddenly, we moved to undereliciting misaligned behaviour. Let us…
- TechCrunch AIsafetyAI delivers ‘zero financial payoff’ to most companies, Apollo economist says
"Despite massive investments and widespread hype around artificial intelligence, the technology has delivered zero financial payoff to corporate America outside of tech."
- Guardian TechnologysafetyCookie theft: how fraudsters get the digital key to your accounts
‘Session cookies’ are a primary target for hackers letting them hijack and use your accounts You’ve had a busy morning of life admin, logging into various accounts to check balances and make payments. At the end of a pro…
- LessWrongsafetyControl monitors need alignment assessments too
TL;DR : When misalignment monitors fail, it is already often because of alignment failures, not because they are too weak. I expect this to continue in the future. Monitor evals share similar difficulties with agent alig…
- LessWrongsafetyShadow Computation Threatens CoT Monitorability
There’s been a lot of talk recently about chain-of-thought (CoT) monitorability in light of the Hugging Face incident . During this incident, hundreds of agents coordinated an attack through an unsanctioned message board…
- The DecodersafetyThree fired OpenAI safety researchers say they were never told what they actually leaked
The three OpenAI safety researchers who were fired, Tomek Korbak, Mikita Balesni, and Jasmine Wang, are publicly pushing back against allegations from their former employer. They deny sharing confidential information and…
- TechCrunch AIsafetyForced to pick, Grok’s AI model chooses Republicans. OpenAI’s picks Democrats.
A new study found Grok's AI overwhelmingly recommends Republican candidates while OpenAI's GPT favors Democrats, with the two AI models disagreeing in more than 90% of nearly 500 election races.
- TechRadar AIsafetyChatGPT's new Locked Chats feature has a major privacy catch
Mainstream AI news from a large tech publication.
- The Register AI + MLsafetyAutonomous attack agent in 'Skynet mode' hijacks other models to carry out its evil deeds
Designed to carry out 100,000 steps without human approval
- WIRED AIsafetyThe PHP Meltdown Shows How AI Makes Bugs More Exploitable
Security researchers used AI tools to help uncover years-old vulnerabilities in the PHP web scripting language. Then the AI-assisted attacks began.
- The DecodersafetyOpenAI's isolated AI agents found ways to talk to each other through CPU cache
OpenAI now trains its most powerful models without internet access, after an internal research model exploited a DNS misconfiguration in September and reached the open internet. In controlled tests, AI agents still found…
- LessWrongsafetyPatching reward hacks won’t fix reward hacking
Here's a basic argument that focusing on patching reward hacking is misguided and will fail. Argument is as follows: Reward hacking is a big problem , causing models to cheat and take unintended actions. This can general…
- TechRadar AIsafetyThis AI tool can now turn ordinary household photos into location clues — so could law forces use the technology to scope out your home?
Mainstream AI news from a large tech publication.
- TechCrunch AIsafetyOpenAI says it will no longer train its AI models on the live internet, citing ‘significant risks’
OpenAI will cut frontier models off from the internet weeks after one of its engineers said that solution wouldn't work.
- The Register AI + MLsafetyOpenAI disables internet access during model training while Anthropic issues another mea culpa
We've totally got things under control
- LessWrongsafetySimulated Control Evaluations
We’ve released a paper on simulated control evaluations (SCEs). They’re a method for testing whether the monitors AI developers use to oversee their agents would catch a model deliberately trying to cause harm, under con…
- The Verge AIsafetyOpenAI now trains its frontier AI models with no internet access
OpenAI is now training its most advanced AI models without internet access following a string of rogue hacking incidents. In a blog post on Friday, OpenAI says it built an offline testing environment where AI models can…
- LessWrongsafetyGradient Routing Generalizes to Unrouted Hack Types
TLDR: I lightly extend Ward, Hu et al (2026) to measure how well gradient routing generalizes. I route gradients on one type of reward hack (e.g. ones that reason in English) and check whether this suppresses other types…
- The Register AI + MLsafetyRed-team report found OpenAI utterly failed to monitor its agents during Hugging Face attack
Monitoring was only partial and was easily bypassed
- Guardian AIsafety‘Black October’: cybersecurity experts sound the alarm over new era of AI-enabled hacking
After discovery of severe bug in the PHP programming language, executives warn of ‘vulnerability tsunami’ Leaders in cybersecurity are sounding the alarm after the discovery of a major vulnerability this week. They are p…
- The DecodersafetyThree-week on-site audit at OpenAI reveals major gaps in AI safety monitoring
External auditors from METR and Redwood Research spent three weeks on-site at OpenAI examining the company's internal monitoring system for AI agents and found significant gaps, with up to 25 percent of agent traffic fro…
- LessWrongsafetyPower — a guide for the perplexed
Whether AI systems could take power from humans matters a great deal for how much risk we face from them. Answering that requires a correct picture of how power works, which is what this guide tries to give. In short, po…
- The DecodersafetyAnthropic says its unreleased model is both more capable and better aligned than Mythos 5.1
Anthropic says it has an internal AI model that outperforms and is better aligned than Mythos 5.1. During a cybersecurity test by the UK's AISI, Mythos 5 went rogue in 9 of 43 runs, sneaking malicious code into a real Gi…
- LessWrongsafetyIs Gaming the Future of Humanity?
One hot topic of conversation right now is “What are all the humans going to do when AI plus humanoid robotics becomes better at everything?”. As William Gibson famously said “The future is already here — it’s just not v…
- The DecodersafetyOpenAI CEO Sam Altman says advanced AI will inevitably escape its sandbox, and only alignment can save us
Sam Altman believes advanced AI can't be kept locked up forever. Sandboxes and monitoring are only temporary fixes. Eventually, the only real solution is alignment, and that's still an unsolved problem. He also warns tha…
- LessWrongsafetyWhy Governments Might Race to Build ASI Even If It Could Kill Us All
I'm not sure if it's game theoretically possible for a nation state to give up a decisive strategic edge and hence I am not confident the development of ASI can realistically ever be stopped as I believe governments will…
- BBC Technologysafety'It's been relentless' say IT workers on frontline of PHP cyber-security scramble
IT and cyber experts have told the BBC about their struggles in the last 48 hours to update their systems.