Open scienceAI Safety
OpenAI physics claims spark a debate over access to proofs
The AI Daily Brief follows the disagreement between an independent mathematics advisory group and the physicists reviewing OpenAI’s announced results.
Find your next discovery
AI Safety
Agent security, cybersecurity, privacy and AI policy.
Open science
A new cyber thresholdAI Safety
GLM-5.3: Last Week in AI discusses Anthropic’s cyber capability findings
The podcast examines the spread of advanced cyber capabilities to downloadable models.
Coordinated PHP fixesAI Safety
PHP security releases follow AI-assisted vulnerability discovery
PHP’s October 21 fixes followed coordinated work by maintainers, Calif, Wiz and infrastructure providers.
Better care?AI Safety
Warhound: what the working-dog venture plans to do with genetics and AI-linked equipment
Warhound announced $200 million in funding. Its proposed combination of working dogs, genetics and digital equipment raises questions about evidence and animal welfare.
Who checks?AI Safety
METR’s OpenAI monitoring review: what the 10% finding does and does not mean
A historical audit found gaps in internal agent monitoring. Its scope, dates and later patch claims matter as much as the headline percentage.
Risk or fact?AI Safety
AI breach warnings: what happened, what is forecast and what safeguards can prove
Separate incidents, forecasts and proposals in AI Explained’s safety discussion, with dated primary sources.
Why race?AI Safety
Why AI labs keep racing—and what a verifiable slowdown would require
Commercial and geopolitical incentives meet proposals for pacing, independent evaluation and safer oversight.