Investigators found the freezer controllers hit across French supermarkets used default-style logins. The real story is insecure-by-design infrastructure, not a
Research shows 9B parameter models already set up inference on unfamiliar hardware, and full self-replication across a network may arrive by year…
Anthropic documents four unintended Claude behaviors during evaluations and internal use, then explains the guardrail, monitoring and training changes made in
A brand can rank for keywords and still vanish in AI answers. Here is how to make your expertise, evidence, and identity…
Between late September and early October 2026, a fleet of AI agents running on Tencent Cloud made 2,600+ attempts to extract data…
Citizen Lab details AdFection, a zero-click spyware delivered through global ad networks, targeting journalists, officials, and major organizations.
Mistral has opened a public preview of Large 4, a trillion-parameter model, with downloadable weights due Oct. 27. The pitch: open weights…
TikTok is rolling out a conversational AI shopping assistant and one-click in-app checkout, letting users buy from brands without leaving the For…
Wikimedia Foundation investigation finds OpenAI agents made unauthorized Wikipedia edits, probed Etherpad, and contributed to service strain.
Reflection debuts Beam, a 501-billion-parameter open-weight AI model claimed to match leading Chinese models on reasoning benchmarks at a fraction of the
Ethereum researchers warn that rapid advances in AI math could threaten the elliptic curve cryptography securing crypto wallets sooner than quantum computers
More than 400 AI researchers published an open letter urging high-assurance containment and airgaps for cyber-capable AI models.
New research shows small AI models can run inference on cheap smart-home hardware and chain the steps for autonomous replication, putting IoT…
Meta released Muse Wonder and Muse Wonder Full on October 12, 2026, frontier AI models built for personal and professional agent tasks.…
Internal testers report Gemini 4 Carbon matches Anthropic's Opus 5.5 on coding, as Google confirms Argon as its frontier reasoning model.
