Red Teaming stories

Anthropic and OpenAI take rival paths on AI cyber tools, as one keeps access tightly restricted while the other widens vetted user access.

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

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

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.

Anthropic limits Claude Mythos to critical users after it exposed browser and Linux flaws, stoking fears of AI-driven cyber risk for banks.

LangWatch releases open-source AI red-teaming framework to expose hidden vulnerabilities in production agents through multi-turn attack simulations.

Anthropic is to give UK banks controlled access to its Mythos AI model, as financial firms brace for a new era of autonomous cyber threats.

Netskope's Tony Burnside warns AI agents are creating hidden east-west traffic, calling for omni-directional controls and smarter DLP to stop data leaks.

Anthropic rolls out Claude Opus 4.7 with sharper coding, stronger image handling and new cyber safeguards, plus fresh API controls and review tools.

Testlio rolls out human-led AI chatbot testing as brands face pressure to prove safety, accuracy and trust before customer use.

AI testing lags as more than half of organisations ship features, with 52% saying fewer than half of projects reach full production.

Applause appoints Aatish Salvi as Chief Technology Officer to steer AI-driven testing as the software quality specialist expands its enterprise offering.

Synack launches AI-driven assessment to expose overlooked attack surface gaps as offensive tools speed up vulnerability discovery.

Abacus secures CREST accreditation for penetration testing, bolstering its pitch to regulated sectors as demand rises for verified cyber security assurance.

Anthropic’s Claude Code is under scrutiny after researchers found deny rules can weaken in long workflows, raising fresh concerns for AI-driven development.

ChatGPT flaw may have let attackers siphon sensitive user data via DNS queries, prompting OpenAI to issue a fix after researchers exposed the bug.

F5 and Forcepoint have teamed up to link data discovery with runtime controls, aiming to curb AI risks as enterprises move systems into production.

F5 and Forcepoint team up to give enterprises continuous AI security, linking data discovery with runtime controls to reduce risk in production systems.

OpenAI opens public Safety Bug Bounty to find agentic prompt injection, data exfiltration and other AI misuse risks.

CrowdStrike broadens its consumption-based model into services, offering partners and new customers flexible access to incident response and advisory support.