AI Safety stories

Lovelace's Elemental aims to give enterprise AI verifiable evidence, using knowledge graphs and cited outputs for high-stakes decisions.

Cloud Security Alliance deepens agentic AI oversight with catastrophic risk controls, CVE authority and two governance specs.

AI-enabled medical devices are deepening healthcare cyber risk as hospitals adopt them faster than governance, security controls and legacy systems can cope.

SANS says AI has become routine in the most dangerous cyber attacks, leaving defenders racing to keep pace with faster, smarter intrusions.

Denodo's Dominic Sartorio warns businesses accelerating artificial intelligence deployments that cloud migrations alone do not make data ready for real-time AI.

Cequence Security adds AI Gateway controls to curb autonomous agents' access to enterprise apps and data with persona-based permissions.

Anthropic's Agent Harness tests how much businesses should trust AI agents, as experts warn automation could amplify failures and cyber risk.

Jazz and Aptech call for stronger IP safeguards as AI and video-first content increase exposure risks and reshape digital creation.

Check Point and Flashpoint warn AI is speeding up exploit development, shrinking patch windows and forcing defenders to rethink first-line protections.

Ping Identity-backed research says enterprises need stronger runtime controls as autonomous AI agents expose gaps in authorisation and oversight.

SAS has unveiled AI Navigator, a governance service that gives organisations a single view of AI models and agents as scrutiny intensifies.

Google Cloud's Thomas Kurian unveils Gemini Enterprise as an open, secure workplace AI platform linking models, data, chips and security.

Netskope deepens Google Cloud tie-up to bring AI Guardrails to generative and agentic systems with real-time safety checks.

Bitdefender urges journalists to harden phones, accounts and source contacts against phishing, spyware, impersonation and AI-linked leaks.

Anthropic's Claude Mythos cyber security model and Project Glasswing are fuelling fears that AI-driven vulnerability hunting is becoming a boardroom issue.

BAND secures USD $17 million seed backing to build an interaction layer for multi-agent AI systems, as firms seek better oversight and interoperability.

AI tools are speeding up vulnerability hunting and could help lower-skilled hackers exploit flaws far sooner, Flashpoint has warned.

Zapier broadens enterprise AI controls with policy enforcement across workflows, agents, assistants and SDK-built apps.

Isaca launches AI risk certification as European research finds many organisations still lack clear ownership, response plans and board oversight.

CrowdStrike unveils AI security coalition with Accenture, EY, IBM Cybersecurity Services, Kroll and OpenAI to spot and fix code flaws faster.