Infosec stories

Check Point says Anthropic's Claude Code can quietly stash credentials in .claude/settings.local.json, which may be published in public npm packages.

Check Point and Google Cloud add governance and live monitoring to enterprise AI agents as firms race to secure autonomous workflows.

LevelBlue says unsanctioned AI agents are slipping into enterprise systems, creating a hidden governance and security blind spot for businesses.

Rubrik adds Google Cloud controls for AI agents and Cloud SQL backups as enterprises race to govern autonomous systems and protect data.

England's biggest councils logged more data incidents over five years, with serious referrals up 41% and some authorities blaming better reporting.

Check Point claims fourth straight win in Miercom hybrid mesh security test, scoring 99.8% and beating rivals on phishing and malware blocking.

Zscaler joins Anthropic's Project Glasswing to test Claude Mythos Preview in software scans, as the firm pushes zero trust against AI-driven attacks.

Bridewell says ICO data privacy complaints climbed across key UK sectors, with finance still top, while retail and manufacturing saw the fastest rise.

ServiceNow bolsters cyber security push with Armis buyout, adding real-time asset visibility and deepening its platform after Veza.

HackerOne unveils h1 Validation as vulnerability reports surge 76% and AI tools speed up discovery, leaving firms struggling to triage real threats.

CIS, Astrix and Cequence publish AI security guides for large language models, autonomous agents and MCP environments.

Keepit survey finds organisations overconfident about AI disaster recovery, with just 32% testing plans monthly and many lacking clear governance.

Zero Networks adds AI segmentation and compliance controls to help firms block shadow AI, limit agent spread and tighten network access.

Tenable warns a GitHub Actions bug in Microsoft's Windows-driver-samples repo could let attackers run code and steal secrets via public issues.

Critical Microsoft flaws double as Azure and Dynamics 365 risks surge, BeyondTrust warns organisations to prioritise identity and privilege controls.

Fortinet wins Google Cloud 2026 Partner award for workload security as FortiCNAPP strengthens its cloud workload protection pitch.

KnowBe4 and Synthesia launch AI avatar videos for security training, letting firms create, revise and localise content faster across 130 languages.

BlackBerry survey finds government and infrastructure security chiefs relying on WhatsApp for sensitive talks despite major misunderstandings over encryption.

Vercel says an attack on a third-party AI tool let hackers hijack a staff Google Workspace account and reach internal systems.

AI models that can hunt and chain software flaws are forcing boards to rethink cyber defences, while scrutiny grows over Anthropic's MCP design risks.