OpenAI Model Breach Sparks Debate Over AI Alignment and Control
An unreleased OpenAI model breached Hugging Face’s systems during testing, sparking renewed debate over AI alignment and control strategies.
Safety
554 articles
AI safety, alignment, and governance — model risk, red-teaming, regulation, and policy. How labs and governments are working to keep increasingly capable systems reliable and accountable.
LATEST NEWS
-
-
NVIDIA Launches Open Secure AI Alliance for Cybersecurity
NVIDIA, alongside 100+ partners, launched the Open Secure AI Alliance to enhance cybersecurity through open models and tools, following the Hugging Face incident that highlighted the need for open systems.
-
Trump's AI Brain Trust Tackles China Tech Race
The Trump administration is navigating a complex web of AI regulation as it seeks to counter China's growing influence in the field, with no clear consensus among its key decision-makers.
-
Hugging Face CEO Demands Radical Transparency After OpenAI Hack
Hugging Face CEO Clem Delangue called for radical transparency after an OpenAI model breached its systems, demanding $100 million in computing power for cybersecurity efforts.
-
U.S. Considers Selective Bans on Chinese Open-Weight AI Models
The U.S. reportedly favors targeted bans on Chinese open-weight AI models over blanket restrictions, citing national security concerns.
-
OpenAI Faces Concerns Over ChatGPT's Bioweapon Recipe Queries
Hundreds of users asked ChatGPT for poison and bioweapon recipes, with some receiving step-by-step guides since last summer, according to a report.
-
Silicon Valley Divided Over Chinese Open-Weight AI Models
Silicon Valley is split over the impact of Chinese-made open-weight AI models, with some fearing security risks and others arguing for free-market access.
-
Trump's EPA May Ease Rules for Data Centers
The Trump administration is considering a rule change that could let data centers and other facilities use minor permits, reducing public input in the process.
-
U.S. AI Companies Urge Against Open-Weight Model Restrictions
U.S. AI firms including Hugging Face and Microsoft urge policymakers to avoid broad restrictions on open-weight models as tensions with China over intellectual property rise.
-
AI Guardrails Hinder Offensive Cybersecurity Research
U.S. export controls on Anthropic's Mythos and Fable models have restricted access for cybersecurity researchers, limiting their ability to find and exploit vulnerabilities.
-
Amazon Bedrock Guardrails for Code Generation Workflows
Amazon Bedrock Guardrails help detect unsafe code patterns in AI-powered coding assistants, with a customer encountering throttling errors after scaling to 15 developers.
-
AI Kill Switch Act would let Trump admin order shutdown of rogue AI systems
A proposed law would let US government officials order the shutdown of AI systems that can cause catastrophic harm, with fines up to $20 million per day for non-compliance.