AI Safety stories
Attackers are using agentic tools to compress breaches into days, exposing developers and cloud users to faster, harder-to-stop intrusions.
Businesses face a tougher test for AI agents as DevRev's new benchmark measures accuracy, cost and access controls in enterprise settings.
Enterprise buyers risk signing off on AI systems that only claim human oversight, while real-time intervention and auditability are often absent.
Developers can now isolate AI-generated code and user scripts inside Cloud Run, reducing the risk of exposing credentials or host data.
The model can run for up to an hour at 720p and 60 frames per second, with real-time controls and multi-user interaction.
ChatGPT users worldwide will get more natural spoken exchanges, as the new full-duplex system listens and talks at the same time.
Businesses could see faster, cheaper AI coding and office workflows as OpenAI rolls out GPT-5.6 with stronger safeguards.
AI is helping hospitals cut scan times, clear backlogs and spot disease earlier, while doctors still keep final say over treatment.
Business and public sector organisations faced 2,270 attacks a week in June, as ransomware rose 33% and GenAI use exposed sensitive data.
Businesses can now build AI agents more cheaply, as the open stack matches top closed models on one benchmark while cutting run costs sharply.
Enterprise buyers are demanding proof that AI agents can be audited, tested and constrained before they go live in customer service.
Rising use of deepfakes and voice cloning is forcing firms to rethink staff training as insurers and buyers scrutinise human risk more closely.
Weak default safeguards and uneven sandboxing could leave developers exposed to command execution before workspace trust is granted.
Community banks and credit unions could get faster access to specialist AI tools as Eltropy opens its platform to vetted fintech builders.
Regulated industries may get a safer route to production AI as the tie-up offers tighter control over data, governance and deployment.
The controlled trial could help security teams cut false positives and speed remediation as frontier AI moves beyond finding bugs to validating risk.
Nearly half of businesses have paused or scaled back AI projects as weak cost tracking leaves returns unproven and security gaps widen.
Australian users can now automate office tasks across apps and files as OpenAI rolls out ChatGPT Work and GPT-5.6 nationwide.
The bank says the platform is already resolving more enquiries end to end, as it replaces legacy systems across 2 million monthly conversations.
It aims to curb fragmentation as businesses test autonomous AI agents that need to verify identity, access data and prove compliance.
