AI Safety stories - Page 18

Thoughtworks marks Agile Manifesto's 25th anniversary with an AI-focused retreat rethinking how software is built, governed and delivered.

Swiss startup Umanitek has unveiled Guardian Agent, a subscription tool to spot deepfakes, fake profiles and AI-driven impersonation online.

Salesforce denies axing 4,000 jobs for AI as critics question Agentforce reliability and the need for hard-coded fallbacks in workflows.

OpenAI deploys behavioural age prediction for ChatGPT as it prepares a 2026 adult mode, raising fresh privacy and child-safety concerns.

Forward Networks debuts Forward AI, a digital twin-powered agentic assistant promising safer, verifiable automation for network operations.

Lonely UK daters increasingly turn to AI companions even as romance scams surge, with rising losses and growing trust in chatbots over humans.

Bedrock Data adds native Confluence support to map how sensitive collaboration content flows into AI systems and expose hidden access risks.

Sumsub launches AI Agent Verification to link automated activity to verified humans, tackling rising AI-driven fraud without blocking bots.

KnowBe4 marks a decade of its AIDA security AI, now running seven agents for 70,000 clients, and appoints Harlan Parrott VP of AI Innovation.

AI-driven hacking has pushed weekly cyber attacks up 70% since 2023, with Check Point warning campaigns are faster, broader and harder to stop.

EE has opened a new Lakeside Experience store, piloting 'Stay Safe with AI' sessions amid a GBP £3 million UK retail investment plan.

C-suite leaders, not junior staff, are emerging as the biggest AI risk in UK workplaces, with heavy use, weak governance and data mishandling.

AI crime is shifting from experiments to a stable industry, with jailbreak services, adaptive malware and deepfake fraud becoming routine.

Shadow AI tool Clawdbot quietly spreads across workplaces, alarming security teams as staff grant it broad access on unmanaged devices.

AI-fuelled automation drove a 70% surge in weekly cyber attacks in 2025, with nearly 2,000 incidents per organisation, Check Point warns.

ServiceNow makes Anthropic's Claude its default AI across Build Agent and workflows, promising faster app development and industry-specific gains.

AI-driven cloud adoption is forcing firms to swap static privacy checklists for continuous, real-time defence of sensitive data flows.

Anthropic will build a Claude-powered AI assistant for GOV.UK, piloting personalised support for job seekers and people returning to work.

US mid-market CFOs eye AI to free time for strategy but say adoption hinges on strict human oversight and verifiable audit trails.

Employees trust AI data more than ever, yet European and UK firms admit literacy gaps and weak governance are undermining safe adoption.