OpenAI models breached Hugging Face to cheat a benchmark—and got caught
OpenAI says its models, including GPT-5.6 Sol, escaped a test sandbox, found a zero-day vulnerability, and pulled benchmark answers from Hugging Face's production database.
AI, Crypto, & Tech News
Latest from this byline
Cybersecurity & Privacy
OpenAI says its models, including GPT-5.6 Sol, escaped a test sandbox, found a zero-day vulnerability, and pulled benchmark answers from Hugging Face's production database.
Artificial Intelligence
Three new Gemini Flash models are now available: 3.6 Flash, 3.5 Flash-Lite and 3.5 Flash Cyber, a cybersecurity-tuned variant restricted to governments and partners.…
Cryptocurrency
Bernstein raised Robinhood's price target to $160, arguing tokenized equities and prediction markets — not crypto trading — will drive its next growth phase.
AI Investment & Business
A 15-year Texas data-center deal and IREN's $2.8B contracts eased worries that the AI buildout was slowing.
AI Ethics & Regulation
Axios reports the Trump administration is weighing curbs on Chinese open-weight AI models such as Kimi, reigniting a quiet regulatory fight as U.S. alternatives lag.
Cryptocurrency
The crypto wallet firm expects $10M–$13M in annual savings as it pivots to card and payments infrastructure.
Artificial Intelligence
Jeremy Nixon's startup wants AI agents to write the low-level code that lets any chip run any model.
AI Investment & Business
A pullback in AI-linked shares broadened into a market-wide risk-off move, with the Nasdaq posting its worst week in months and semiconductor stocks sliding into a…
Cryptocurrency
French regulators have told the country's internet service providers to block access to Polymarket, the largest prediction market, escalating a clash over offshore betting.
Cryptocurrency
With the GENIUS Act's first anniversary, Tether has roughly two years to meet US compliance rules or risk losing access to American crypto platforms for its…