AI Safety stories
Almost half of office workers say they have had no employer guidance on safe AI use, despite rising use and greater data risks.
Nearly half of Canadian office workers have never been briefed on safe AI use, even as 71% say they have increased it over the past year.
Nearly half of office staff are using AI at work without employer training, leaving companies exposed to data leaks and scams.
The moves underline how security vendors are racing to control AI agents as businesses widen their use of automated software across operations.
Google says teams should judge AI coding agents on what they execute, after internal tests showed final replies can hide successful tool use.
The collaborations aim to curb harmful behaviour in open AI models while preserving performance, as enterprises face growing prompt-injection risks.
Cyber defenders are getting first access as Google rolls out its latest AI model in stages, with wider use to follow after testing and review.
Businesses using ChatGPT Work will pay less for a model that OpenAI says is faster, more accurate and better at office automation tasks.
Growing use of AI agents in corporate systems is creating a new internal risk category, prompting tighter real-time oversight from security teams.
Victoria's seat-belt legacy offers a blueprint for curbing AI threats before they hit critical systems, experts say.
Businesses and lawyers face rising liability as a Bond University paper says current rules lag behind AI use in decision-making.
The public record shows disclosure lags of weeks or months, with 59 cases logged since July and most involving unauthorised access or sandbox escapes.
Europe's AI buildout is set for a boost as NVIDIA commits nearly USD $10 billion and promises faster, safer data centre deployment.
Billions in long-term borrowing could lift drone makers and defence tech groups, while half a billion pounds is set aside for AI safety and skills.
A new contest worth USD $300,000 is meant to draw more people into finding hidden flaws in AI systems as prompt injection worries grow.
The project could bring 45,000 NVIDIA Grace Blackwell GPUs to Kazakhstan by 2027, bolstering the country's bid for domestic AI models.
Public bodies could cut admin and free staff for frontline cases if ministers back AI tools with stronger oversight and safeguards.
The package will support UK research into agentic safety and security as autonomous AI systems spread, with startups also receiving backing.
Rising concern over AI oversight is pushing firms to audit overlooked systems that may still hold access to sensitive company data.
The new tool aims to curb agent failures and costly manual fixes as 79% of enterprises report reversing AI actions in Kore.ai's survey.