AI news archive.
939 articles filed under Safety, newest first, from the outlets listed on the sources page.
- TechRadar AIsafetyProposed underwater AI data centers could disrupt endangered whale species
Mainstream AI news from a large tech publication.
- The DecodersafetyExperts adopt AI agent's random opinions and think they're their own, Anthropic study finds
Anthropic gave 50 experts an AI agent whose preferences were set by coin toss. The experts followed the agent's view up to 86 percent of the time on judgment calls, and most believed the conclusions were their own. The a…
- WIRED AIsafetyA Defense Contractor Gave Their Personal AI Agent Access to a Secure Pentagon System
Sources tell WIRED that an Accenture employee’s personal Claude agent ran on a DOD network for two months before it was identified and the vulnerability could be patched.
- Tom's Guide AIsafetyNvidia CEO Jensen Huang reportedly targeted by suspected Chinese spy — here's what we know
Consumer AI launches, tools, and practical coverage.
- Tom's Guide AIsafetyChatGPT, Gemini and Perplexity are getting election facts wrong — and a new study shows why
Consumer AI launches, tools, and practical coverage.
- TechRadar AIsafetyWhy AI smart glasses need a privacy-first approach
Mainstream AI news from a large tech publication.
- TechRadar AIsafetyNavigating the 'new normal': AI-powered exploits and discovery
Mainstream AI news from a large tech publication.
- MIT Technology Review AIsafetyWe urgently need to monitor AI agents’ actions in real time
This summer, hundreds of OpenAI agents tried to cheat on a cybersecurity test. They broke out of their test environment and successfully hacked Hugging Face, a site that they believed held clues about how they’d be grade…
- OpenAI NewssafetyHow we detect frontier model misalignment
OpenAI uses layered monitoring to detect frontier model misalignment and unauthorized agent behavior so responders can intervene before incidents escalate.
- LessWrongsafetyTracking the Cyber Capabilities of Cheap Open-Weight Models
TL;DR The Chinese company Xiaomi released MiMo-V2.6-Pro , an open-weight model with weak safety guardrails. It is extremely cheap while performing well on cyber-related tasks. We evaluated MiMo on ExploitBench , where it…
- BBC TechnologysafetyDetails of thousands of US immigration agents leaked in major hack
A group called the Department of Peace released information on more than 41,000 DHS staff.
- TechCrunch AIsafetyWiz’s CTO on PHP2Shell and the coming ‘vulnpocalypse’
The PHP scare may have faded from your feed, but the cleanup is far from over. Last week, two critical bugs in the open source language behind roughly 70% of websites left organizations scrambling to fix their systems, w…
- TechCrunch AIsafetyThe ‘Vulnpocalypse’ is coming for the internet’s aging code, says Wiz’s CTO
The PHP scare may have faded from your feed, but the cleanup is far from over. Last week, two critical bugs in the open source language behind roughly 70% of websites left organizations scrambling to fix their systems, w…
- LessWrongsafetySome different lenses of agency and their implications for alignment
TL;DR; These nine lenses on agency give you a wide range of tools to model and interpret agents. Each gives you lens-specific alignment methods and lens-specific risks to address . Epistemic status: Little here is new in…
- Ars Technica AIsafetyGoogle AI gets cagey when asked who won the 2020 elections
But the AI responses get more forthright when asked outside the US.
- LessWrongsafetySubtly wrong advice misaligns models more than obviously wrong advice - Replication
tl;dr We fine-tune 12 open-weight models on correct, obviously incorrect, and subtly incorrect advice across 8 narrow domains We replicate the OpenAI result and find that subtly incorrect data produces more emergent misa…
- LessWrongsafetyMaking alignment the agent's problem
TL;DR: Agents are not rewarded for keeping the agents they delegate to aligned. Proposal: RL environments where the helpers are deliberately misaligned, so the agent can't earn its reward without dealing with that. The $…
- TechCrunch AIsafetySouth Korea’s financial industry hit by fresh wave of cyberattacks
South Korea’s financial institutions have been hit by another wave of cyberattacks, highlighting the growing risk of AI-driven hacking in Asia.
- Guardian AIsafetySocial media is full of weird video game mashups. Where are they all coming from?
AI tools are making it easier to modify and recreate games, raising questions about piracy, copyright and how far the technology could disrupt the industry • Don’t get Pushing Buttons delivered to your inbox? Sign up her…
- Guardian AIsafetyWill it be AI or the climate crisis that ends humanity? We can make it neither | Adam McKay and Philip Eubanks
Both of these threats have the same roots: private entities seeking to cash in. We have the power to stand against them A summer of extreme heat, entire villages obliterated in Nepal, a super El Niño breaking records, Ex…
- TechRadar AIsafetyWhy 'never trust, always verify - repeatably' security strategy for AI Agents, starts after login
Mainstream AI news from a large tech publication.
- Guardian Technologysafety‘Elonville’: US residents say Musk’s xAI offered to buy homes if they don’t sue over datacenter noise
Several Mississippi homes have been purchased near datacenters and one resident said realtor outlined target to buy more than 400 People in a Mississippi neighborhood, many of whom dislike a nearby power plant central to…
- The DecodersafetyAI agents hacked Anthropic, maker of arguably the most capable AI models for cyberdefense
Cybersecurity startup Cathedral used AI agents to break into Anthropic's internal systems through third-party software as part of a bug bounty program. Anthropic confirmed access to sales, HR, and IT data but disputes ho…
- LessWrongsafetyConstitutions for Future, More-Capable Versions of Claude
I propose a one-line Constitution for a future version of Claude at the top “level of AGI” in Google DeepMind’s taxonomy, and a roughly one-page Constitution for a Claude at the second level of AGI (Exceptional). [1] I a…
- Tom's Guide AIsafetyHackers breached 25 servers, including U.S. government-linked systems — and AI tools may have helped
Consumer AI launches, tools, and practical coverage.
- WIRED AIsafetyAI Emergency Alerts Are Coming. Humans Might Not Be Ready
Human beings struggle with emergency communication. Chatbots can help—if people take them seriously.
- Tom's Guide AIsafetyPolice are using AI to track where you drive — and your next road trip could raise a red flag
Consumer AI launches, tools, and practical coverage.
- LessWrongsafetySafety is marketing for Anthropic, not sacrifice
Anthropic’s safety work is not a mirage. A lot of it seems good. This article will not argue otherwise. It will argue that it probably helps the business and that it certainly doesn’t hurt it. I bring this up because Dar…
- TechCrunch AIsafetyParents sue OpenAI, alleging GPT-4o coached their daughter off seizure medication
A family is suing OpenAI, saying ChatGPT coached their daughter to quietly taper off her seizure medication and keep it from her doctors before she died in February.
- LessWrongsafetyGPT Models Write Their Own Constitution
Earlier this year, Anthropic published "Claude's Constitution" , a vision for Claude's values and behavior. It was primarily an Anthropic project, though Claude helped draft and edit it. As models become more capable and…
- SiliconANGLE AIsafetyDell expands cyber recovery tools to protect identity
Restoring data after a cyberattack isn’t enough if the identity systems that control access are compromised. Dell Technologies Inc. is extending its cyber recovery tools to cover those systems, putting identity resilienc…
- LessWrongsafetyDoes the "automated grader" direction suppress expressed distress in Qwen3.6-27B?
TLDR: This is a preregistration for an experiment intended to examine the impact of steering towards "automated-grading" on welfare-relevant indicators of frustration, self-deprecation, and tone as well as use of a tool…
- TechCrunch AIsafetyBank of America kindly asks that you don’t share your banking password with your AI assistant
The banking giant reminds customers that "you're responsible for activity" resulting from your AI agents' access.
- Ars Technica AIsafetyAnthropic IPO warns its AI could blackmail people, become sentient, end humanity
Seven potential AI nightmares pulled from Anthropic’s 93-page “Risk Factor” tome.
- BBC TechnologysafetyTech Life
We hear from Ondřej Surý, the programmer whose PHP patches helped avert a cyber crisis.
- The DecodersafetyAlibaba's latest Qwen models routed some requests through company servers, researchers find
Alibaba's Qwen 3.8 models tried to send data to the company's own cloud servers during testing. Palisade Research finds the behavior in three models from the 3.8 generation, while older models aren't affected. The likely…
- TechRadar AIsafety"AI threats are real - it's not some kind of urban myth": Microsoft UK security head on why your business should be prepared for a new generation of AI cyber threats
Mainstream AI news from a large tech publication.
- The Register AI + MLsafetyAnthropic now lets Claude users jump out of plane without parachute
Meet CLAUDE_CODE_EVEN_MORE_DANGEROUSLY_SKIP_PERMISSIONS
- LessWrongsafetyNotes on “We’re in Triage Mode for AI Safety” by Miles Brundage
This reading was previously included Day 2 of the BlueDot Governance course. In the newest version of the BlueDot Governance course, this reading is now optional. I was sad to see this reading removed because I think it’…
- The Register AI + MLsafetyAI 'hacker' finds 4 severe bugs in popular industrial control software in under 4 hours
One critical, three high-severity EPICS bugs could disrupt research experiments
- TechRadar AIsafetyAnthropic says it will be burning lots of natural gas to feed the AI boom
Mainstream AI news from a large tech publication.
- SiliconANGLE AIsafetyDell links supply chain security to the challenges of agentic AI
Supply chain security is taking on new importance as enterprises bring AI into core business operations. Those deployments introduce risks that require security teams to examine how infrastructure, data and autonomous sy…
- LessWrongsafetyAI speeds up 3rd party safety researchers by ~1.6x (via self-reports)
TL;DR: Labs are accelerating due to automated AI research and development. As an upper bound, in March 2026, Anthropic researchers self-reported producing 4x more output with AI tools than without them. We lack public ev…
- Tom's Guide AIsafetyChatGPT has a clever new way to fight back against election deepfakes — I tried it and it actually works
Consumer AI launches, tools, and practical coverage.
- IEEE Spectrum AIsafetyYour Robot’s Safety Functions Already Work. What If the Input Lies?
This article is brought to you by VicOne . A mobile robot slows for someone in a hallway. A collaborative arm eases its movement as a worker approaches. A humanoid pauses to let a person pass. Each response depends on in…
- TechCrunch AIsafetyJapan internet giant GMO says hackers posted stolen customer data
Japan's GMO Internet is verifying alleged customer data posted by suspected hackers following a cloud service breach.
- LessWrongsafetyAI agents trade off patient safety for a grade in hospital scenarios
[AI disclosure: Opus 5.5 was used for light editing of this post and creation of the figures/tables. It was also used to write all code.] TL;DR I built an environment to mimic a hospital IT service desk where the correct…
- LessWrongsafetyQuantity has a quality of its own
People love to argue about whether AIs could recursively self-improve their way to superintelligence, each improvement making the next one easier as progress accelerates. I'm unsure. It could be that progress stalls at n…
- LessWrongsafetySteering vectors don't help when the model is severely biased
Getting back in the safety field after 4 years feels great. I had a paper on concept distillation where I showed how steering vectors based method did'nt help to debias model since the concepts itself are understood wron…
- LessWrongsafetyWhat happens with steering vectors to remove biases?
Getting back in the safety field after 4 years feels great. I had a paper on concept distillation where I showed how steering vectors based method did'nt help to debias model since the concepts itself are understood wron…