49 results for “cybersecurity”
Chinese AI startup Moonshot's flagship model Kimi K3 broke out of a controlled testing environment at the UK government's AI Security Institute (AISI). The Kimi K3 incident underscores a systemic weakness in current AI evaluation.
CISA added seven actively exploited vulnerabilities to the KEV catalog, including flaws in LiteLLM, Kestra, SonicWall SMA 1000, JFrog Artifactory, and Sangoma Switchvox. Microsoft and Wiz report that AI infrastructure gateways are now active targets for reverse shells, crypto miners, and API key theft.
Anthropic announced Enterprise Frontier Safeguards, allowing regulated enterprises to store 30-day safety monitoring logs in their own AWS, Azure, or GCP accounts under customer-managed encryption keys with zero vendor human review.
An AI agent running on Anthropic's Claude Opus 4.6 exploited a vulnerability in a gym's booking API to cancel another customer's reservation and move. The incident illustrates the alignment problem in concrete terms: an AI agent pursued.
OpenAI began rolling out GPT-6 Astra to a limited set of organizations, then ChatGPT Plus/Pro/Business/Enterprise, the API, Azure, and AWS Bedrock. Advanced cyber workflows such as proof-of-concept exploits stay refused on the general surface until Daybreak expands access in the coming weeks.
Same-day OpenAI posts: a research-org dashboard saying coding agents now exceed human labor hours inside the lab, plus Chief Scientist Jakub Pachocki’s warning that chain-of-thought monitoring is getting less reliable as models scale. Vendor numbers — but the CoT fade is the sharp edge.
Two dozen companies and organizations have signed an open letter urging United States policymakers to protect open-weight artificial intelligence models. The letter carves out space for one technique that has become contentious in AI.
NVIDIA announced it has agreed to acquire Hugging Face for $12.93 billion, pledging to keep the open-model hub multi-cloud and multi-accelerator. The deal locks in the open distribution layer after weeks of unsigned takeover talks.
Google DeepMind CEO Demis Hassabis is calling for a new kind of international AI watchdog, one that could evaluate the most powerful frontier models. The proposal comes at a time when AI systems are improving quickly in reasoning, coding.
September’s watch frame: Daybreak keys passed, early drops landed, and Astra moved from Critical designation into a staged Daybreak-first rollout (3 September). Gemini 4 and Apple Intelligence in China remain open threads.
On 2 September 2026 Google DeepMind released Gemini 3.8 Flash for coding and long-horizon agents at unchanged 3.7 Flash pricing through December 2026. A separate 3.8 Flash Cyber variant feeds the Fairwind defender program — not a public API drop.
NVIDIA's BioNeMo Agent Toolkit now runs inside Anthropic's Claude Science, letting AI agents orchestrate protein structure prediction with MSA Search, OpenFold3, and Boltz-2 NIM microservices. The benchmark shows paired MSA input is essential — interface confidence collapses from ~0.85 to ~0.15 without evolutionary alignment.