Anthropic’s New AI Threat Report is Absolutely Insane
Chinese labs ran 200 million exchanges through Claude. A weapons cell in Yemen used it as an engineering team.
Zeniteq is now available on mobile for iPhone, iPad and Apple Watch.Download nowLearn more
Explore AI news, practical guides, tutorials, reviews, and insights from Zeniteq.
Chinese labs ran 200 million exchanges through Claude. A weapons cell in Yemen used it as an engineering team.
The new nonprofit will pursue earlier threat detection and faster medical countermeasure development, backed by public funding and up to £200 million…
A criminal threat actor used an AI model to discover and weaponize a 2FA bypass before Google intervened and got the flaw…
Private negotiations over publication rights expose a gap between promises of independent scrutiny and the terms governing what evaluators can disclose.
Google’s threat researchers describe an autonomous patch-to-exploit pipeline that could shrink the time defenders have to secure exposed systems.
CNN reports that a false chatbot analysis of a ship’s cargo reached military planners before a last-minute review stopped an interception.
Executives urged caution at Dublin’s AI summit, but their claims leave business buyers needing evidence about specific models, deployments and threats.
The reported preparations raise a harder question: how much do public safety commitments reveal about efforts to prevent a severe incident?
Former lab employees warned of weak oversight as New York City considered third-party validation, incident reporting, whistleblower protections and civil liability.
A reported safety-policy trigger and a new coalition statement sharpen the dispute over Anthropic’s push to automate AI development.
A controlled Qwen3 experiment shows how poisoned pretraining data can produce a command-execution backdoor that survives clean agent training.
The reported proposal ties catastrophic-risk thresholds to independent evaluations, a 72-hour reporting window and a court order, while leaving important questions about…
House leaders are threatening to compel Jensen Huang’s appearance while two senators seek preservation of government records on Nvidia exports and reported…
Nasdaq suffers its worst crash since April 2025.
The Palantir chief spots a real incentive problem, but liability alone cannot answer the catastrophic risks Anthropic describes.
A cloud-provider compromise reached Fireworks’ development systems, exposing customer data and showing why AI retention promises need to be checked across every…
The reported discussions have no disclosed price or timetable, but an existing open-source security partnership helps explain the potential fit.
Leaked chats show Aur0ra operators reframed live intrusions as authorized simulations, exposing a basic authorization weakness in coding-agent guardrails.
Jacob Coxon helped train the AI models. Then he quit and told 90 million people they should be scared.
Four nonprofits want a California judge to halt new AI model development and deployment at five companies, but no injunction has been…
Censys found publicly exposed ARTEX consoles, including systems hosting attack tools, but the count is an exposure finding, not a tally of…
His new essay calls for AI taxes, human-reserved work, stronger safety rules, and U.S.-China coordination before disruption outruns policy.