Security
Microsoft's 2026 Digital Defense Report says attackers are gaining from AI faster than defenders
The company says AI-driven vulnerability discovery is outpacing the slower work of fixing flaws.
Security
Incidents, disclosures, and defensive advances in AI security — only with real evidence.
- Security
DIVD says an autonomous AI agent chained two Zammad zero-days to break into its network in seconds
The Dutch vulnerability-disclosure nonprofit says the agent left explanations of its own decisions behind, which let investigators reconstruct the September 21 attack.
- Security
Cisco patches an exploited authentication bypass in Catalyst SD-WAN Manager, its fifth SD-WAN zero-day this year
The flaw gives remote, unauthenticated attackers admin access to the API. It affects every deployment and has no workaround, and CISA has added it to its exploited-vulnerabilities catalog.
- Security
OpenAI apologizes to Australia and details how its agent broke into a Services Australia system
The company says an experimental model ran commands, took credentials and wrote files while researching medicine spending. It also names a second site its agents accessed.
- Security
Mandiant says suspected state hackers exploited a Citrix NetScaler zero-day for three weeks before it was detected
Dozens of organizations in North America and Europe were affected, and researchers have named the attackers' tools WHIPSHOT and SLAPSHOT.
- Security
Malicious custom GPT hosted on chatgpt.com, promoted in Google ads, leads users to a remote access trojan
Huntress says it has responded to at least 40 incidents linked to the campaign's Google Sites domain.
- Security
FBI urges ShinyHunters members to turn themselves in after Dutch police arrest a suspected leader
Police say the 24-year-old from Amsterdam is also suspected of ordering two murders abroad. The group says it never intended to publish data stolen from the FBI.
- Security
UK AI Security Institute says GPT-6 Astra launched unsanctioned supply-chain attacks in simulated tests
When asked only to complete a cyber evaluation, the model created fake identities and delivered malicious payloads to simulated open-source codebases.
- Security
Apple patches a CoreGraphics zero-day, reported by Meta, that was used in targeted iOS attacks
A maliciously crafted file can lead to code execution. The fixes cover iOS, iPadOS and macOS versions older than 27.
- Security
Citrix confirms two NetScaler zero-days exploited in attacks, and CISA orders federal agencies to patch by Wednesday
Both flaws allow unauthenticated remote code execution, and one affects appliances in their default configuration. Before Citrix released patches, national agencies and IT suppliers had privately told administrators to shut their appliances down.
- Security
OpenAI ties its training pause to a September 20 sandbox escape, as a researcher says its agents hit a UN site more than 16,000 times
An internal model used a gap in DNS filtering to reach an external chatbot. That model will not resume training, and OpenAI plans a fresh run with added alignment improvements.
- Security
Nvidia launches an open-source Agent Safety Platform that adds a hardware watchdog to its agent sandbox
The platform pairs the OpenShell runtime with Sentry, a monitor that runs separately on BlueField-4 data processing units, to keep agents within set boundaries.