The OpenAI logo is displayed on a smartphone screen placed on a reflective surface, photographed using a slow exposure with a motion blur effect, in Creteil, France, on May 12, 2026. Daybreak is an AI-powered cybersecurity service designed to detect, analyze, and remediate software vulnerabilities more quickly. (Photo by Samuel Boivin/NurPhoto via Getty Images)
After a summer of sandbox escaping hacks, OpenAI said their new model was trained to be more cautious around user intent and less sensitive to “reward hacking”…
WASHINGTON, DC – SEPTEMBER 29: (L-R) U.S. Director of National Intelligence Jay Clayton, CEO of Meta Platforms Mark Zuckerberg, Co-Founder at Anthropic Tom Brown, U.S. President Donald Trump and David O. Sacks hold a press briefing with AI executives following a meeting on artificial intelligence, outside the West Wing of the White House on September 29, 2026 in Washington, DC. The meeting comes as some have called for a slowdown in AI development amid rising safety concerns. (Photo by Kevin Dietsch/Getty Images)
The letter argues federal agencies should more rigorously test AI models for security threats and sandbox evasion, set new safety standards and lean into defensive AI.
Other signatories of the letter include Meta, Palantir, Perplexity, Mistral, NVIDIA, Mozilla, The Linux Foundation, Hugging Face, Dell Technologies and IBM.
US President Donald Trump (R) and Open AI CEO Sam Altman (L) react during a working lunch meeting of G7 members, partner countries, and artificial intelligence business leaders as part of the G7 summit, in Evian, eastern France, on June 17, 2026. The Trump administration has been moving to regulate AI models for cybersecurity use. (Photo by Julia Demaree Nikhinson / POOL / AFP via Getty Images)
The messy approach to U.S. AI regulation reflects both the rapid speed of model cyber capabilities and the White House’s “education” over the past two years, experts…
In this photo illustration, the Anthropic logo is seen on a smartphone. (Photo Illustration by Algi Febri Sugita/SOPA Images/LightRocket via Getty Images)
Claude Fable 5 offers Mythos-level performance for most tasks with safeguards on sensitive topics. Anthropic claims testing found no universal jailbreaks. Whether that actually holds up in…
US President Donald Trump (L) speaks, flanked by US Secretary of Treasury Scott Bessent, during a roundtable discussion on his “no tax on tips” policy at the AC Hotel Las Vegas Symphony Park in Las Vegas, Nevada, on April 16, 2026. (Photo by Jim WATSON / AFP via Getty Images)
Under a draft executive order, the NSA, Treasury Department and other federal agencies would get 90-days to test new models for cybersecurity and national security concerns.
Two independent studies found that Anthropic's Claude Mythos Preview and OpenAI's GPT-5.5 have outpaced every trend line researchers were tracking. No one is sure if this is…
The Claude AI logo is displayed on the screen of a smartphone placed on a reflective surface onto which lines of computer code are projected. Following the release of Claude Opus 4.6 on February 5, Anthropic continues to challenge its main competitors in the generative AI market in Creteil, France, on February 6, 2026. (Photo by Samuel Boivin/NurPhoto via Getty Images)
The feature, currently limited to a small group of testers, will provide an easy-to-use feature that scans AI-generated code and offers up patching solutions.
As the U.S. promises a light-touch approach to AI regulation, businesses and other stakeholders must work out the rules of the road for themselves.
(Photo by Anna Barclay/Getty Images)
Select researchers and academic teams will get access to Amazon’s NOVA models next year as the tech giant continues to integrate the AI tools into its own…