AI Safety stories
Companies adopting AI agents in payments now have a new way to spot compliance and revenue risks before customers are affected.
Eligible research organisations can now access OpenAI's updated GPT-Rosalind model, as the company widens its life sciences rollout worldwide.
The new feature targets shadow AI on laptops and desktops, helping security teams block data leaks before models can access sensitive files.
The tie-up keeps analytics and AI inside Snowflake's security boundary, while reducing mismatched business definitions across dashboards and agents.
Frontline teams at multi-location service businesses could cut dashboard churn as AskNicely's new tools automate insights and routine review replies.
The partnership could speed up flaw detection and patching for critical software used by businesses and public sector organisations across the region.
Enterprise security teams face a new visibility gap as approved AI agents can copy and transfer sensitive data in under 30 minutes.
As AI agents multiply in Snowflake environments, security teams need tighter controls to stop machine-speed access from outpacing policy.
Growing pressure to prove AI decisions is pushing manufacturers towards tighter governance, connected data and MCP-based integration by 2026.
Developers and enterprise customers will get more AI controls as Microsoft adds agents, in-house models and security tools across its software stack.
The new method could make multimodal AI outputs easier to trust in medicine and other high-stakes uses by tying answers to stated reasoning.
Pilot trials suggest the setup could cut factory energy use by 10% and lift assembly-line productivity by 12%.
The models are aimed at developers and enterprises, with Microsoft saying internal training could cut costs and improve control in regulated industries.
The model will first be tested inside Mayo Clinic, with the partners aiming to improve clinical decisions before wider rollout via Azure.
The London-based AI data start-up has won support as demand grows for specialist training datasets for reasoning-heavy models.
The funding will help the Edinburgh fintech expand tools that let banks check AI agents meet conduct standards for customers.
The conference will put Scotland's AI talent, security and infrastructure under the spotlight as debate over governance and control intensifies.
Most Australian workers using AI at work have had no formal training, leaving security, privacy and skills gaps as adoption races ahead.
The survey also found most firms still lack secrets scanning and rapid audit proof, leaving hidden credentials and compliance delays as weak spots.
The filing could sharpen competition in driver monitoring and telematics as insurers and fleet operators seek faster risk and incident alerts.
