AI Safety stories - Page 20

HackerOne unveils Good Faith AI Research Safe Harbor, giving security testers clearer authorisation and protection for probing AI systems.

Hubtel IT boosts staff by 25% and targets GBP £2.5m turnover as it ramps up AI and cyber security services for UK small businesses.

ServiceNow expands its OpenAI partnership to make frontier models and voice-first AI core to enterprise workflows and automation at scale.

Tech Mahindra's sovereign multilingual AI, handling millions of queries in Hindi, Bahasa and regional dialects, is honoured by WEF's MINDS.

Thomson Reuters launches Trust in AI Alliance with Anthropic, AWS, Google Cloud and OpenAI to develop trustworthy agentic AI systems.

Attackers are already exploiting AI agents, extracting hidden prompts, bypassing safety checks and abusing tools tied to data and systems.

Synechron has launched an Agentic AI suite to automate mission-critical workflows in heavily regulated financial and insurance sectors.

The Isle of Man has unveiled a National AI Office, backed by GBP £1m, to steer responsible AI use across government and the wider economy.

Anthropic is setting up an Australian office in Sydney as Claude adoption soars, anchoring a wider global expansion and hiring push.

Ben Goertzel and Jaron Lanier clash over whether increasingly autonomous AI deserves moral standing or must always remain under human control.

A deepfake surge is driving relentless AI-powered celebrity scams, with persona kits and adaptive fraud campaigns eroding online trust.

Ofcom has opened a probe into X over claims its Grok AI tool enabled deepfake sexual images, testing the UK's new Online Safety Act.

BioCatch warns AI shopping agents will turbocharge online fraud, urging banks and retailers to distinguish helpful bots from criminal misuse.

Cloudflare snaps up Human Native in bid to reshape how AI models pay for online content, promising new tools for pricing and discovery.

Quali adds an Agentic Control Plane to Torque, promising real-time governance for autonomous AI workloads and GPU-heavy infrastructure.

F5 rolls out AI Guardrails and AI Red Team to harden runtime security, blending adversarial testing with real-time policy enforcement.

OpenText warns 2026 will bring an AI‑driven identity crisis, deepfake-fuelled scams and decision overload that many firms are not ready for.

Agentic AI promises rapid software gains, but a recent Claude Code cyberattack shows how “double agents” can outpace unprepared developers.

Nintex warns enterprise agentic AI will face a 2026 reality check, with tighter budgets, targeted use cases and tougher governance.

Progress predicts 2026 will mark a shift from AI pilots to orchestrated, governed platforms, widening the gap between leaders and laggards.