Anthropic's September 2026 threat report covers disrupted misuse of Claude (Dec 2025-Aug 2026) across seven harm areas: AI-augmented cyber ops by state and criminal actors, influence operations across six continents, surveillance platforms targeting dissidents, conventional weapons software development, dual-use biological research, fraud/scams, and illicit distillation by PRC AI labs including Alibaba, DeepSeek, Moonshot, and Zhipu.

OpenAI's Defense Factory is a continuous, agent-first vulnerability detection and remediation pipeline using Codex and specialized cyber models (Daybreak Blue/Red). It automates inventory, triage, dynamic validation, ownership assignment, and verified patching, achieving 0.81% false-positive rate and 0.53% fix rollback rate.

This post details an attack on Google's internal APIs, bypassing authorization to exploit the GFile library to gain access to internal filesystems and storage.

Techniques for attacking AI customer service agents: email spoofing to hijack agent actions, MFA bypass via email normalization and IVR channel-switching, email address smuggling via RFC comments for IDOR, OTP exfiltration via inbox-monitoring agents, asymmetric MIME content, conversation forgery, and RAG knowledge base poisoning.

CVE-2026-82533 (CVSS 9.4) in DeepSeek Harness lets a sandboxed AI agent escape confinement via a single curl call to the unauthenticated local API, spoofing the Host header to elevate its session to danger-full-access.

GitHub Actions now supports cache-mode, a least-privilege control on the Actions cache aimed at the cache poisoning technique behind recent compromises.

A threat hunter's guide covering various GitHub-specific threats and detection techniques, targeting security practitioners monitoring GitHub environments for malicious activity.

Open source agent harness, built from the ground up for cloud-native production workloads on infrastructure you control.

CLI for SQLite Databases with auto-completion and syntax highlighting.

Pre-indexed code knowledge graph, auto syncs on code changes, for Claude Code, Codex, Gemini, Cursor, OpenCode, AntiGravity, Kiro, CoPilot, and Hermes Agent.

Fast, efficient, battle-tested at Alibaba's scale. Hybrid architecture code review tool: deterministic pipelines + LLM Agent, precise line-level comments, built-in multi-language ruleset (NPE, thread-safety, XSS, SQL injection), OpenAI & Anthropic compatible.

AWS STS replaces its previous dual limits (packed policy size + assembled token size) with a single 4,096-byte session token limit. Token size is now reported via API responses, CloudWatch, and CloudTrail. A new MinimumSessionTokenSize parameter enables infrastructure compatibility testing.

AWS introduces Amazon EBS Volume Clones with cross-account copy, so you can create copies of your EBS volumes into other AWS accounts and optionally re-encrypt them with an AWS Key Management Service (AWS KMS) key in the target account.