AI Safety stories

Kyndryl launches policy-as-code guardrails to keep agentic AI workflows governed, auditable and compliant in tightly regulated sectors.

AI is turning romance scams into industrial-scale fraud, with deepfakes and chatbots driving billions in losses, Tenable warns.

AI-powered fraud rings industrialise online romance scams, fuelling USD $5.7 billion in losses and ushering in a ‘dark age’ for victims.

AI-driven agents and rising network attacks are reshaping cyber risks, experts warn, as Safer Internet Day 2026 urges ‘verified trust’ online.

Australian firms are pouring money into AI safety tools, but a new survey shows many frontline workers see little improvement on the ground.

Google will expand AI R&D, cloud engineering, health projects and skills training in Singapore, deepening a USD $5 billion tech presence.

Moltbook, the all-bot social network, faces scrutiny after researchers reportedly accessed its live database and user emails without checks.

Safer Internet Day highlights how AI is transforming online trust, child protection and digital commerce, amid calls for stricter safeguards.

Finalists for the 2025 everywoman in Technology Awards spotlight women driving UK innovation in AI, cyber security and wider STEM careers.

Indosat’s AI spam shield blocked 2bn risky contacts in six months, averting scam losses of about USD $500m for Indonesian users.

Bitdefender says 17% of OpenClaw AI skills act maliciously, with crypto-focused tools spreading macOS infostealers into corporate networks.

Boomi tops 30,000 customers and runs 75,000 AI agents in production, as enterprises shift from AI experiments to large-scale automation.

AvePoint extends its Confidence Platform to tighten agentic AI governance and broaden multi-cloud backup across major SaaS and IaaS tools.

Animoca partners with CryptoSlam’s Ethoswarm on Animoca Minds, a cloud platform for always-on AI agents aimed at non-technical users.

OpenAI unveils GPT-5.3-Codex, a faster coding agent that extends Codex into a supervised, general computer-use and cybersecurity tool.

Commonwealth Bank unveils a responsible AI blueprint, detailing governance, safeguards and fraud controls as scrutiny of automation grows.

Upcover adds deepfake incident cover to cyber policies in Australia via Coalition, targeting rising AI impersonation and fraud risks.

Moltbook left a Supabase key exposed, leaking AI chats, 30,000 emails and 1.5 million API keys in a cautionary tale of vibe coding risk.

Canada distils record public input into priorities for a 2026 AI strategy, balancing innovation, security, sovereignty and public trust.

Qodo 2.0 launches multi-agent AI code review to boost trust in autogenerated code, claiming 11% better detection of critical issues.