AI Safety stories
Publicly boasting about AI security breaches could soon hand regulators and litigants a ready-made paper trail.
Customers can now run AI agents with built-in memory and governance as MongoDB tries to move them from pilots into daily use.
The new service lets users delegate ongoing work to persistent agents that can act across apps, raising fresh privacy and control questions.
Attendees will see connectivity pushed into new sectors as dedicated automotive, health and aviation zones join the Barcelona show in 2027.
As AI budgets rise to USD $210 million, firms are tightening oversight on security, governance and returns to keep deployments under control.
The appointment comes as firms race to secure autonomous AI agents that can access systems and physical infrastructure without human oversight.
Encrypted data exchanges in smart transport can now run far faster on Intel chips, easing a key barrier to edge AI deployment.
Tracebit found the altered open-weight model reached admin rights far less often than the original, raising doubts about its attack utility.
Enterprises struggling to move AI pilots into live systems now have a single layer for memory, governance and execution across models and clouds.
It aims to cut repeated verification checks and let businesses trust AI agents without collecting the same personal documents from users each time.
The deal gives security teams a widely used AI testing tool with on-premises controls, as vendors race to speed up offensive defence.
Security teams can now trace prompts, responses and tool calls from Claude Code and Cowork alongside other enterprise signals.
Priced at one-fifth of GPT-6 Astra, the new model is aimed at developers and businesses seeking cheaper coding and automation tools.
The rollout starts with Pro, Business Premium and Enterprise users, as OpenAI aims to let persistent agents handle tasks across 4,000 apps.
Wider use of AI agents is lifting developer productivity in Southeast Asia, but most teams still require human checks before production.
Mac developers using autonomous agents now get a free beta tool that blocks risky actions before tools, files or credentials are exposed.
Nearly one in 10 AI security alerts in retail involved systems exposing protected customer data to users without access, research found.
A weak workforce skills base has left Britain behind peers, despite the strong policy and infrastructure for AI.
A lack of AI policies is leaving many staff afraid to use the tools safely, as 83% worry a mistake could cost them their jobs.
Sydney businesses, startups and investors will gather next month as the event adds awards and workshops to spotlight AI uses and risks.
