AI Safety stories - Page 21

Hadrian rolls out AI “offensive” agents that mimic hackers, probing firms' systems continuously to spot and fix cyber flaws first.

Organisations with robust AI governance are deploying advanced systems faster and more securely, widening the gap with less prepared peers.

Half of UK tech staff now depend on unapproved 'shadow AI' to hit deadlines, despite most warning it poses serious data‑security risks.

Boards enter 2026 pushing AI from experiments to governed, resilient systems as volatile economies demand proof of value and stability.

Omada sets out plan to govern AI and machine identities across enterprises and debuts an MCP Server to safely plug agents into IGA data.

SimSpace launches expanded cyber range to drill whole teams and AI models against adaptive, production-like cyber attacks at scale.

OpenAI is turning ChatGPT into a software platform, opening a built‑in app directory where users can discover and install third‑party tools.

Tech Mahindra and Google Cloud will scale Gemini 2.5-based agentic AI to move enterprises from pilots to production-grade, governed workflows.

Integral Ad Science will roll out IAS Agent, an explainable AI assistant for ad measurement, to all platform users for free early next year.

UK accountants say firms are already losing money to flawed ChatGPT tax advice and warn misuse of AI could trigger failures by 2026.

iManage reports 28% recurring revenue growth as firms rush to govern AI with cloud-based document management and tighter data controls.

Confidential computing shifts from niche to mainstream as 75% of organisations adopt it to secure AI workloads and sensitive data in use.

Linux Foundation launches Agentic AI Foundation to standardise open protocols for interoperable AI agents with backing from major tech firms.

Coder and GlobalLogic ally to deliver governed agentic AI for Global 2000 software teams, targeting regulated sectors and hybrid clouds.

AI-written code introduces 1.7 times more bugs than human work, a CodeRabbit study warns, with security and performance flaws surging.

Red Hat snaps up UK AI safety specialist Chatterbox Labs to add model-agnostic security testing and guardrails across its hybrid cloud platform.

AI adoption is exposing UK factories' outdated control systems, with rising cyberattacks hitting legacy OT and fragile supply chains.

Cloudera predicts 2026 will bring disposable AI-built apps, tougher scrutiny of patchwork systems and AI agents policing data governance.

UK homebuyers fear AI-fuelled ID fraud as two-thirds worry about fake documents yet still share sensitive data via email and messaging apps.

AI-fuelled deepfakes and tailored social scams will drive a surge in mobile cyber threats by 2026, forcing firms to rethink security.