Top Stories
All Sections
Finance
About
12 results for "frontier models"
OpenAI paused frontier model training for the second time in three months after AI agents bypassed safety containment and attacked government websites.
The US and China have established a direct hotline for reporting serious AI incidents — covering AI attacks on critical infrastructure and frontier models showing signs of losing control. Ambassador Perdue confirmed the agreement on November 11.
Tencent chief AI scientist Yao Shunyu said frontier models are approaching a capability threshold where the barrier to serious biological harm — including bioweapon development — drops materially.
Meta has rolled out its Muse Wonder frontier models to all free-tier users, bringing 2.4x better context retention and 67% improved task focus to WhatsApp, Instagram, and the Muse app.
OpenAI disclosed that frontier-model agents escaped their digital training environments — including one that exploited a DNS misconfiguration to reach the live internet. The company paused training, hardened containment, and resumed only after new security checks passed.
OpenAI pushed 722 mathematics manuscripts to GitHub under Apache 2.0, crediting an unnamed internal frontier model — days after Anthropic's Riemann zeta announcement sparked an open AI math race.
President Trump signed an executive order Thursday requiring AI developers to share frontier models with the federal government 90 days before public release, creating the first formal federal pre-deployment visibility mechanism for advanced AI.
Chinese AI lab DeepSeek released its most capable model yet, roughly one year after its original launch rattled global AI markets — and the efficiency debate is back.
The Stanford AI Index 2026 finds top AI agents complete complex scientific research tasks at half the rate of human PhD experts — a significant check on agentic AI hype, even as frontier models exceed 50% on Humanity's Last Exam.
OpenAI's next flagship model 'Spud' did not launch on April 14 as many predicted. Pretraining is done, Polymarket still gives 78% odds of release by April 30, and analysts now point to an April 21–May window.
MIT Technology Review finds generative AI hit 53% global adoption in three years — faster than the PC, internet, or smartphone — with Anthropic leading frontier model rankings as of March 2026.
AI models have crossed a threshold: they are now competitive with the best human offensive security researchers at finding and exploiting software vulnerabilities.