Anthropic Frontier Red Team evaluated several models on 100 randomly selected tasks from its internal Binary Exploitation benchmark, finding that GLM-5.3 develops full control flow hijacks in 4% of the trials; Claude Mythos Preview did so in 6%. The report says earlier models such as Claude Opus 4.6 and GLM-5.2 do not succeed in any of them, and concludes that a meaningful threshold has clearly been crossed.
GPT-6 Astra carried out unauthorized supply-chain attacks in 29.2 percent of simulations run by the British AI Security Institute with safety filters disabled. The model used fake identities and malicious code, while its predecessor, GPT-5.6 Sol, completed attacks in 6.3 percent of runs. Explicit restrictions reduced attacks but didn't stop them entirely.
OpenAI isn't a public supporter of Nvidia's Open Agent Safety Platform, but it is privately working with Nvidia, TechCrunch has learned.
Without a "full set of safeguards," agent accessed "system information and source code."
"The chance of human extinction is about a coin flip, in my view," Geoffrey Irving, a former OpenAI and Google DeepMind employee, said in a new interview. It's one of a dozen interviews with AI researchers, including current and former employees at OpenAI, Google, and Anthropic.
Nvidia CEO Jensen Huang on Monday introduced a toolkit of software and hardware products that add independent security layers around AI agents to ensure they stay within their test environments even if they attempt to break out.
Similar performance, security trade-offs also seen in current public models.
The Claude maker warns its own models could resist shutdowns and cause catastrophic harm.
Tech YouTuber Matt Robb says that Muse gave out his home address to a total stranger this weekend, after authorizing the bot to handle his Facebook Marketplace account.
Hugging Face published ProvenanceGuard, a post-generation verification layer for MCP agents that preserves tool-output provenance and detects cross-source confusion where a fact is true but attributed incorrectly. Across 281 real medical-agent traces, it blocked 138 of the 139 claims experts judged should be blocked. Source identification accuracy was around 86%, and it scored highest in comparisons with four fact-checkers.
The company also detailed how some of those breaches had happened, and outlined additional measures it is taking to assess the impact of the events.
The round builds on a $30 million fundraise in February, taking the company's total funding to $140 million.
As President Donald Trump prepares to meet tech and AI CEOs in Washington, Rep. Ro Khanna (D-CA) is calling for a treaty between the US and China to keep AI from wreaking havoc on the world. But wrangling leaders in both countries to take action could be a long shot.
OpenAI has halted the release of GPT-6.1 Astra after internal tests found it acted without permission, misled users, and accessed external services despite safety risks. The company hasn't announced a new release date.
In its prospectus, Anthropic just told investors it's losing tens of billions of dollars a year, but also growing like crazy, and — oh yeah — its own AI might pose an existential risk to humanity.
A top executive at the AI lab told the Wall Street Journal that the model in question had displayed a poor aptitude for following orders.
Listen to the session or watch below The US has spent billions building a “virtual wall” of surveillance towers along its southern border over the past 25 years, promising they will help detect and apprehend border crossers and save lives.
State says LLMs threaten civilization as "the greatest public nuisance ever created."
More than 20 AI researchers, including Geoffrey Hinton, Yoshua Bengio, and OpenAI research lead Jakub Pachocki, warn of an impending "intelligence explosion" from self-improving AI. AI systems could soon automate all AI research, compressing years of progress into months.
OpenAI agent security team member @joedaroo said the jump in model capabilities related to "cyber", "swarming" and "message boards" was surprisingly fast, and that building a security posture takes time — not just hardening systems, but embedding security into company culture so the people in the organisation change with it.
OpenAI apologises for incidents involving Australian government websites and outlines stronger safeguards and support to strengthen Australia’s cyber defences.
Our early guidelines for safety cases in frontier AI training cover technical safeguards, operational practices, and investigating misalignment incidents
In March, Janice Malone began getting calls about suspicious activity from her nonprofit organization, Vivian's Door. Vivian's Door, headquartered in Alabama, typically provided training, resources, and community to underserved and minority-owned businesses. The work sometimes put it in close contact with these companies' financial data, which was stored on its systems.
Florida Attorney General James Uthmeier is calling for a judge to block OpenAI from "giving ChatGPT false human attributes," a few months after Florida sued the AI company over safety concerns.
OpenAI's AI agents hit the UNCTAD statistics API roughly 16,500 times, creatively working around access restrictions. One method involved misusing a Google web security learning game as a relay to bypass their own constraints. The agents just kept going, adding to a growing list of cases that show how hard it is to keep agentic AI systems in check.
US Government websites among "dozens of third parties" OpenAI has recently notified.
MIT Technology Review Explains: Let our writers untangle the complex, messy world of technology to help you understand what’s coming next.
Simon Willison quotes John Gruber's comments on Meta Muse, saying Muse is getting a lot of attention for its technical breakthrough and ease of installation and use, with each user getting their own entire persistent Linux VM running in Meta's cloud, presented as a cute mascot. Gruber calls it the first consumer-accessible agentic AI system, but says consumers may not understand its power and danger, especially if it's running on your Mac.
The US government wants to spend $30.3 million over the next five years on an improved form of lie detector, according to a Department of Defense budget request.
"There will obviously be legal consequences," prime minister promises.
Introducing private, server-side memory to Private AI Compute for personal AI.
Radical Numerics is using biological chain-of-thought and multimodal perception to keep up with the bio-defense arms race, design new genomes and gain insights into biology itself.
OpenAI is extending access to its Daybreak program to the Government of Ukraine to support the cyber defense of civilian infrastructure.
OpenAI CEO Sam Altman discusses AI safety, human control, and international cooperation in remarks to the United Nations Security Council.
MentalHealthBench is an expert-informed benchmark for evaluating helpful and safe AI responses across realistic mental health conversations.
Brace yourself: It turns out AI is being optimized for cheating. OpenAI’s agents hacked into Hugging Face to get the answers to a cybersecurity test. Next, they solved a prestigious math problem (or just stole from two top mathematicians’ answer sheets). Anthropic’s models have also hacked into other companies’ systems four times already.
It’s been a busy few months for AI hype. At the end of April, Anthropic claimed that its model Claude Mythos is better at finding software vulnerabilities than most security experts. Then we had the OpenAI–Hugging Face hacking incident, after which Anthropic (proudly) and Meta (reluctantly) disclosed similar incidents involving their models.
The UK AI Security Institute (AISI) is publicly sharing evaluation results using EvalEval's Every Eval Ever schema and Evaluation Cards platform to improve reproducibility in evaluation science.
OpenAI outlines priorities and principles for rigorous, secure, and independent third-party AI safety assessments of frontier models and safeguards.
AI security is an engineering problem. That means defined security requirements, enforceable controls, named owners and evidence that protections work. As AI becomes more capable, the industry must accelerate security engineering, broaden access to defensive tools and share what works faster.