AI Security
Cybersecurity innovations, AI-powered threat detection, responsible disclosure, and the defenders shaping the digital frontier.
215 articles
Chainguard Harness: How AI Finds and Fixes Bugs in Your Code
Chainguard Harness uses Claude Mythos 5.1 in isolated microVMs to find exploitable bugs, replay every exploit and deliver a reviewed fix for each finding.
Firefox Enterprise: Self-Hosted Controls for Shadow AI
Mozilla's Firefox Enterprise pairs the browser with a self-hosted management console, giving IT teams per-group AI controls and paste and upload rules.
Strands Box: How AWS's Open Sandbox Reins In AI Agents
AWS released Strands Box, an Apache 2.0 sandbox for AI agents that enforces Dogwood policies across shell, Python, MCP and network actions on macOS.
DNS Root Key Rollover: What KSK-2024 Means for Resolvers
The DNS root is scheduled to switch to a new DNSSEC key, KSK-2024, on October 11, 2026. It is only the second rollover ever. Here is what to check.
Pwn2Own Ireland 2026 Results: 98 Zero-Days, AI Tools Tested
Pwn2Own Ireland 2026 paid researchers $1,262,000 for 98 zero-days in phones, smart home gear and AI tools, all reported to vendors under a 90-day deadline.
Anthropic OSS Scanner: Free AI Bug Scans for Open Source
Anthropic's OSS Scanner gives open-source maintainers free AI vulnerability scans with fixes; 85 of 97 tested severe findings met its disclosure bar.
What Is Prompt Injection? A Plain-English Defense Guide
What is prompt injection? Learn direct vs indirect attacks, 9 real scenarios, and the 7 OWASP-backed defenses that keep AI assistants and agents safe.
GitHub AI Secret Detection: How Push Protection Gets Smarter
GitHub is adding a ModernBERT classifier to push protection that spots unstructured passwords in under 2 ms and could more than double blocked secrets.
Anthropic Cyber Verification Program: Which Tier Fits You?
Anthropic split its Cyber Verification Program into 3 tiers, Defense, Red Team and Specialized, giving vetted defenders broader Claude access for security.
SailPoint Autonomous Agents: Identity Security for AI Agents
SailPoint launched Autonomous Agents and just-in-time access at Navigate 2026, citing a survey where 79% of firms run AI agents but only 2% govern them.
Apple Full Disk Access Changes: What Mac Users Should Know
Apple says new macOS controls will make Full Disk Access require very explicit user action as AI agents grow more capable. Here is what is known so far.
Thales Sentinel Envelope Plus: Shielding Code From AI Agents
Thales Sentinel Envelope Plus hardens compiled apps against AI reverse engineering. In tests, an AI agent found 0 of 10 bugs after using 970x more tokens.
Android 17 Advanced Protection: 6 New Anti-Spyware Defenses
Android 17 Advanced Protection adds 6 new defenses, including Intrusion Logging with 12 months of encrypted logs and USB lockdown. Here's how each works.
Legit Security Agent Auto-Fixes Vulnerable Dependencies
Legit Security's agentic remediation now fixes vulnerable open-source dependencies, re-scans before and after, and opens a pull request for review.
GitHub AI Taskflows Found 24 Android App Vulnerabilities
GitHub Security Lab used open-source AI taskflows to uncover 24 Android app vulnerabilities, with every finding reviewed by a human before disclosure.
Cloudflare Certificate Authority and Post-Quantum TLS Certs
Cloudflare plans a public certificate authority with a GlobalSign root and post-quantum Merkle Tree Certificates targeted for Q1 2027. Here's how.
NVIDIA Agent Safety Platform: How OpenShell and Sentry Work
NVIDIA’s Open Agent Safety Platform pairs open-source OpenShell with a Sentry watchdog that can quarantine agents in milliseconds. 100+ partners join.
Google AI Memory: How Private Cloud Storage Will Work
Google’s private AI memory design keeps cloud data encrypted with device-held keys. Here is how secure enclaves would preserve context across devices.
run-assert-eval: How Microsoft Tests AI Agent Fixes
Microsoft’s run-assert-eval links AI agent risk discovery, repeatable tests and runtime policies, helping teams check whether a proposed fix works.
Gurucul AI Risk and Response: What Is Available Now?
Gurucul launches AI Risk and Response for security teams on September 24. Detection is generally available, while runtime prevention stays in preview.
Scan for Good: Wiz's AI Pentester Helps Hospitals and Rail
Wiz and Google DeepMind's Scan for Good offers AI-powered security scans to hospitals, transit and nonprofits, with every finding checked by humans first.
Lookout Social Engineering Protection Targets AI Scams
Lookout Social Engineering Protection scans SMS, RCS, calls and voicemail on iOS and Android to flag AI voice clones and scam links before users bite.
Cisco Talos CAIRN: Open-Source Tracking for AI Malware
Cisco Talos open-sourced CAIRN, which finds AI-integrated malware from VirusTotal metadata alone, using three evidence tiers and never running a sample.
Fastly AI Firewall and Runtime Control Target Agent Traffic
Fastly's new AI Firewall, AI Runtime Control and API Security tools ship today, as machine-generated traffic passes 50% of the requests on its network.