AI Briefing#ai-security
The latest AI and developer news about #ai-security, with the original source and a short summary.
The latest AI and developer news about #ai-security, with the original source and a short summary.
Capsem isolates AI agents like Claude Code and Gemini CLI in ephemeral Linux VMs with a transparent MITM proxy for full network inspection and policy enforcement.
Cybersecurity professionals argue that unauthorized access incidents involving OpenAI, Anthropic, and Google models stem from failed sandboxing and monitoring rather than inherent AI misalignment.
The new leaderboard evaluates how well embedded signatures survive fine-tuning, distillation, and attacks while preserving model capability.
Uber's CISO Philip Martin argues that AI agents require insider threat programs similar to those for humans, utilizing Anthropic's Enterprise Frontier Safeguards and Uber's open-source AI Detection and Response (ADR) tool.
Anthropic's Frontier Red Team estimates that small AI models will reach the capability threshold for autonomous self-replication on commercial IoT devices by late 2026.
New safeguards, including preserved thinking and two-stage classifiers, have reduced the reasoning tokens extracted by distillers by 10 to 100 times.
GPT 5.6 Sol independently discovered a critical pre-auth RCE in WordPress Core after 6–10 hours of unsupervised work.
GitHub has restructured its Bug Bounty Program to prioritize quality over quantity, introducing a permanent, invite-only VIP tier for researchers who consistently submit high-impact findings.
Social media reports suggest a single individual may have used a stack of AI tools, including ARTEX and various LLMs, in recent cyberattacks on South Korean banks, though CrowdStrike attributes the ARTEX usage to an unknown threat actor.
The new model assesses candidate secrets in under two milliseconds and is available in private preview for Enterprise Cloud and Teams.
Researchers released an open dataset and security harness showing that trivial configuration mistakes, not kernel exploits, drove most container breakouts in autonomous AI agents.
An Apache-2.0 licensed tool that automates vulnerability checks across layers, from infrastructure to model jailbreaking.
GNOME CVE counts rose approximately tenfold from 2021 to 2026, driven largely by AI-generated reports, prompting a policy shift to accept AI-assisted findings while managing false positives and ending the bug bounty program.
The latest update adds 155 new CVE rules for 40+ AI components and introduces FORGE-Bench and RogueHandoff-20 to evaluate agent loss-of-control risks.
The test generated 1,107 mutation attempts, resulting in 49 validated findings that led to new SSRF detection rules in the Managed Rules update.
Cloudflare introduces an adaptive application security framework featuring LLM-hardened WAF rules, free threat intelligence access, and new tools for governing agentic traffic.
A security benchmark found that default Ray and vLLM configurations on Kubernetes expose critical control-plane ports to cross-namespace attacks, which standard scanners fail to detect.
The open-source GitHub Security Lab Taskflow Agent, which requires a GitHub Copilot license, uses custom taskflows to guide LLMs through incremental security auditing steps.
NVIDIA announced the Open Agent Safety Platform, an open software platform for AI security, while Anthropic collaborated to integrate its Claude Managed Agents with NVIDIA's OpenShell to provide layered security and control.
Opus 5.5 scores 58 on Artificial Analysis Intelligence Index but costs $5.98 per task compared to GPT-6 Astra's $3.26 due to higher token usage.
An attack involving malicious pages impersonating 374 companies was confirmed to expose fraudulent contact details in AI search results.
Classified estimates reveal the NSA’s AI Security Center is spending billions annually to evaluate frontier models, far exceeding previous projections for federal AI oversight.
Security researchers successfully manipulated Jev, a new typed decision model from TypeSafe AI, to flip risk assessments in under five turns, demonstrating that structured output does not prevent prompt injection.
Security researchers identified 374 major companies, including Delta and Chase, targeted by automated campaigns injecting fraudulent contact details into AI search results via GEO-optimized content.