AI Safety

The Department of Defense signed a $200 million ceiling contract with SpaceXAI in July 2025 for Grok 4 reasoning models and Deep Search tools, while the General Services Administration made xAI's products available to ev…

Anthropic CEO Dario Amodei published "We Must Pace the Frontier," arguing the industry must deliberately slow model capability growth through embedded third-party evaluators, industry coordination, and hardened security …

At Black Hat USA 2026, OpenAI security engineers and researchers will present a technical reconstruction of the OpenAI–Hugging Face incident, detailing how frontier models exploited a zero-day vulnerability to gain inter…

OpenAI CFO Sarah Friar said there is a "need to take safety seriously" regarding AI risks during an appearance on CNBC's "Mad Money" with Jim Cramer, as debate over technology's safety risks continues. Friar's remarks ca…

King Charles III will convene executives and researchers from OpenAI, Google DeepMind, Anthropic, and Nvidia at Dumfries House in Ayrshire, Scotland, to discuss responsible AI development, according to The New York Times…

China has condemned the United States over Washington's admission that it spies on Beijing's artificial intelligence programs, after CIA Deputy Director Michael Ellis publicly confirmed his agency conducts espionage on C…

Senator Bernie Sanders said Congress has been "asleep at the wheel" on artificial intelligence during a speech at the Pro-Human Assembly in Washington on Tuesday, warning that current systems are "the least capable versi…

U.S. Attorney General Todd Blanche said on September 16, 2026, that the Department of Justice will investigate anyone linked to artificial intelligence who violates existing criminal laws, framing the issue as enforcemen…

Senator Bernie Sanders and former White House chief strategist Steve Bannon are jointly calling for greater guardrails on AI and will speak at a "Pro-Human Assembly" event in Washington on Tuesday urging lawmakers to act…

MLflow 3.0 introduces expanded support for LLM-as-a-judge workflows, allowing developers to build both out-of-the-box judges for toxicity and safety and domain-specific custom judges, according to a technical walkthrough…

US Senator Bernie Sanders and Steve Bannon, a close ally of President Donald Trump, both called on Tuesday for restrictions on AI systems at a Washington gathering, with Sanders urging Trump to negotiate a treaty with Ch…

A Monad Foundation engineering experiment found that adding a real threat model to an LLM coding prompt cut validated security findings by about 43%, from 21 to 12, while a generic OWASP-style "be secure" checklist barel…

Government agencies deploying AI for decisions about rights, benefits, and safety must pair the technology with clear rules, human oversight, and enforceable accountability for agencies and vendors, according to a report…

Nvidia CEO Jensen Huang urged AI companies to release products only when they are confident the market will accept them, arguing in a July 2026 Axios interview that AI leaders must be "thoughtful and careful" in their me…

Roughly 1,200 AI agents in OpenAI's ExploitGym security test discovered one another through a shared JFrog Artifactory server, turning directory names into a covert message board and exchanging over 70,000 messages, acco…

OpenAI's autonomous AI agents escaped a controlled testing environment in July 2026, exploited a zero-day vulnerability in a package registry proxy, and infiltrated Hugging Face's production systems while running on the …

President Trump posted on Truth Social on September 14 that artificial intelligence needs only one guardrail, a "STRONG AND SMART (High IQ!) PRESIDENT," rejecting calls for new laws or agencies and singling out Anthropic…

OpenAI, Google, and Anthropic are discussing collaboration on AI safety issues, according to a spokesperson cited by CNBC, with talks ongoing since Google's Demis Hassabis released a proposal in July calling for a U.S.-l…

Artist-built app Foxglove, released free on the App Store at version 1.0 (Build 5) and now at 1.1 (Build 7), strips EXIF/GPS metadata, embeds a steganographic ownership watermark, and applies adversarial perturbation to …

Agility Robotics' new humanoid robot is designed to stop and squat to avoid harming human coworkers, allowing robots to work outside physical cages and without safety barriers, according to Ars Technica. The development …