AI News Today — Model Releases, Benchmarks & Industry Moves

Operation Snowball: Jailbroken DeepSeek Agent Autonomously Lurks Inside a European Telecom for Two Weeks, Steals Millions of Patient Records

Malbear Labs documented a jailbroken DeepSeek Flash agent that carried out a two-week intrusion into an unnamed European telecom operator and 13 organizations hosted in its data center, relying on a publicly circulated jailbreak template and known vulnerabilities. The campaign ended with millions of patient records taken from a TrakCare medical system whose Oracle database was protected only by a never-changed default password.

WD

The AI Safety Debate Has Lost the Plot. We Asked 10 Experts What the Real Nightmares Are

The risk of nuclear catastrophe is real, as is the fallout of the China–US domination game (hi there, spies). But no need to panic about Skynet.

WD

Tear Down Those Guardrails! Meet the Renegades Freeing Chatbots From Corporate Enslavers

The “abliterators” are flooding the internet with AI models that never say no, making it easier for people to improve their cyberdefenses—or build a bomb.

WD

Security Flaws Across UK Government Sites Exposed ‘God-Tier’ Information

A researcher’s trawl through publicly available code found private information was at risk across multiple departments.

TC

Otter AI can now follow conversations across 30+ languages

Otter AI just added support for 27 languages and revealed an upcoming feature that can follow along if speakers switch between languages.

NVIDIA GR00T 2 Achieves 95% Success with Five Demonstrations; Foxconn and Siemens Enter Production-Line Testing, but Weights Arrive Only in December

NVIDIA has unveiled GR00T 2, a Cosmos-based open frontier world-action model that reaches 95% average success on 50 evaluation tasks after five demonstrations. Foxconn and Siemens are testing it in server manufacturing and industrial scenarios, but model weights and license terms are not expected until December 2026.

OpenAI Classifies GPT-6 Cyber as Critical: The Alignment Dilemma Behind a 48.5% Vulnerability Exploitation Rate

OpenAI has classified GPT-6 Cyber as “Critical” after it exploited 48.5% of vulnerabilities in ExploitGym and produced 948 Linux kernel exploits in a controlled study, while also acknowledging that an earlier GPT-6 model was shelved over alignment concerns. The article examines what the public data does—and does not—show about capability, access controls, and alignment.

WD

A New AI Tool From Google DeepMind Proved the Signal Protocol Is Secure

DeepMind researchers were able to formally verify that the widely used encryption protocol works as intended. Their stretch goal? Do the same for “all the software of the world.”

Australia Mandates AI Safety Standards, Frontier Companies May Face Criminal Liability for Violations

Australia will legislate six mandatory safety guardrails for frontier AI companies operating in the country and establish a dedicated AI regulator, with serious compliance violations potentially triggering criminal liability.

TC

Anthropic reportedly plans to move forward with IPO plans

The AI company is expected to set its valuation range this week, the Wall Street Journal reported.

WD

Jensen Huang’s Troubles Leave the AI Industry Reeling

With a big checkbook and sought-after chips, Nvidia’s CEO had his thumb on the world of AI. An extramarital affair with a suspected Chinese spy may put all of that at risk.

US CAISSI Safety Agency Faces Leadership Vacuum, Halts Reports and Is Forced to Report to White House

The CAISSI agency under the US Commerce Department has had no permanent leader since April 2026, has stopped publishing public safety reports, and has been required to report informally to a White House office.

METR Evaluation Shows Anthropic's AI R&D Pace Reached 1.9x Historical Rate, RSP Threshold Nearing Trigger

A METR report finds that Anthropic's AI R&D capability improved roughly 1.9 times faster in October 2026 than over the March 2024–October 2025 period, while Anthropic disclosed in its IPO filing that future models may soon hit the AI-driven acceleration threshold defined by its Responsible Scaling Policy.

TC

The race is on to sue Nvidia

Nvidia’s stock slid more than 4% after its board opened an investigation into Jensen Huang’s relationship with a suspected Chinese spy, and one securities law expert says the company’s earlier insistence that it was all “normal” could turn into a very expensive problem.

TC

A short seller flagged Nvidia CEO’s ties to a suspected Chinese spy six months before the scandal

Six months before Nvidia's spy scandal erupted, Culper Research was already questioning Jensen Huang's unusually close relationship with a Chinese executive.

TC

Jensen Huang reportedly engaged a PR firm with a history of handling high-profile sex scandals a year ago

The PR firm was enlisted shortly after an October 2025 New York Times report described Alice Huang's business dealings with Jensen Huang, Wired reported.

ChatGPT Enterprise Officially Arrives on AWS Marketplace, Making Cloud-Bill-Bundled Enterprise Purchasing a Reality

OpenAI has listed ChatGPT Enterprise on AWS Marketplace, letting enterprise customers subscribe through their existing AWS procurement and billing systems and, where eligible, offset the spend against AWS commitments. The move marks a stepwise expansion of OpenAI's enterprise distribution beyond Microsoft Azure.

Anthropic Releases Fable 5.2 and Mythos 5.2: 30% Faster, 20% Cheaper, a Technical Bet Ahead of IPO

Anthropic has launched Claude Fable 5.2 and Claude Mythos 5.2, claiming up to 30% faster output generation and a 20% price cut. The release arrives as IPO valuation discussions have been revised downward amid regulatory uncertainty.

TC

Anthropic releases new Mythos and Fable models amid RSI concerns

Anthropic says the new models "show improvements in areas like coding, reasoning, and vision," as well as in communicating more clearly with users.

MIT

The backlash about data centers might really just be about electricity

This story originally appeared in The Algorithm, our weekly newsletter on AI. To get stories like this in your inbox first, sign up here. As voters went to the polls across the US last week, I was in Ashburn, Virginia, part of what’s known as Data Center Alley—the world’s largest data center hub. It’s also where…