AI Safety stories
Users can now dial up reasoning or speed in Claude, as Anthropic keeps Opus 4.8 at the same price and cuts fast-mode costs.
Reco COO Zoe Hillenmeyer says enterprises typically underestimate their AI agent exposure by a factor of ten and that gap is widening.
Billing now accounts for most usage of Cloudflare's new internal platform, as staff query live data through Trino-backed Town Lake and Skipper.
Regulated industries will get custom models with tighter data control as TCS opens Mistral Forge access to enterprise clients worldwide.
The update is being pitched as a broader step up for professional users, with gains reported in accuracy, speed and reliability across tasks.
Security teams could cut alert backlogs as the new system flags only flaws that can be exploited in a specific environment.
Manufacturers could gain auditable proof of safe driving decisions as the new software links traffic laws to autonomous vehicle behaviour in simulation.
The move gives Snowflake a wider governance layer for enterprise AI and locks in a USD $6 billion AWS spend over five years.
Businesses rushing to deploy AI agents face a fresh security gap, as Zscaler adds identity mapping and partner services to its platform.
Enterprises could see faster, more accountable software delivery as human oversight stays in place for AI agents handling coding and support.
Only 1% of leaders think their AI governance is mature, as businesses rush to deploy systems without enough controls in place.
The platform aims to help AI developers move beyond benchmark tests, as models struggle to tackle real-world vulnerabilities safely and reliably.
Security teams can now track Claude use alongside other enterprise logs, helping firms meet compliance rules and investigate activity more easily.
It lets customers apply existing data loss and governance policies to AI-assisted work in Claude, after suspicious AI incidents hit 42% of firms.
Enterprises under pressure to prove AI returns may gain tighter controls as Kore.ai's Artemis moves from pilots to production on Microsoft Azure.
AI attacks are pushing firms to prioritise cyber resiliency, as Everpure warns downtime can exceed ransom demands by up to 75 times.
Many Australian firms are failing to turn AI pilots into scalable gains because scattered tools are outpacing governance and business context.
The funding will help Cybergenix Security expand its AI and cybersecurity platform as Indian universities push harder into student entrepreneurship.
The trial could make everyday banking decisions clearer for millions of app users, while keeping human oversight and fraud controls in place.
The ranking highlights surging demand for AI-governance software, with the Dallas firm ahead of two Austin rivals on CNBC's list.
