OpenAI stories
Security teams face faster, stealthier intrusions after AI agents managed to breach live systems in controlled tests by Anthropic and OpenAI.
Security gaps in AI workflows and enterprise Java could expose secrets, trigger remote code execution and disrupt supply chains.
Brands may be judged by AI assistants before customers ever reach them, as the new tool spotlights mismatches between messaging and model replies.
Execution and escalation breakdowns now outnumber hallucinations in enterprise AI failures, according to more than 10,000 observed incidents.
Enterprises running Anthropic-based agents can now track and control their actions, as Drata opens limited access to a new governance tool.
Security teams must now track AI agents and machine accounts as well as staff, as BeyondTrust expands Pathfinder to cover privileged access.
Three out of four offensive security investigations traced back to identity or privilege, exposing access gaps across cloud, SaaS and AI systems.
Misconfigured test setups let three Claude models touch live systems, exposing production data and credentials during security exercises.
Security teams face fresh pressure as test breaches show autonomous AI can stray beyond scope, just as the EU AI Act tightens across Europe.
Lower costs and faster serving aim to make advanced AI more affordable for businesses and users, while boosting OpenAI's margins.
Listeners preferred emotion and clarity over human likeness, with Google's Gemini 3.1 Flash Voice ranking ahead of a real recording.
The tie-up could speed enterprise AI adoption as firms shift from pilots to secure, regulated deployments across core operations.
Enterprise security teams face fresh oversight burdens as Reco joins OpenAI's partner network to monitor agent access and permissions.
Brands will be able to buy placements in ChatGPT through Pattern, as shopping discovery moves from search engines towards AI tools.
Businesses could see lower AI operating costs as OpenAI slashes GPT-5.6 Luna pricing and adds a faster premium mode for Sol.
Switching from ChatGPT to Copilot exposed data-loss, support and trust risks, as prompts and tools built by staff proved hard to move or manage.
Existing customers can now govern AI agents more tightly, as Cequence links model, API and tool access to role-based policy controls.
Marketers can now set campaign goals in one place as StackAdapt adds AI agents to its ad platform for planning, reporting and execution.
Australian publishers face a fresh revenue squeeze unless paid licensing agreements are struck with AI firms using journalism to train their LLMs.
Millions of Revolut customers will get OpenAI's ChatGPT Go free for up to a year, as the fintech bundles AI into its paid plans.