AI Midnight
·
History
·
Weekly reports
·
The five risk categories
·
Methodology
·
Sources
·
Glossary
·
Why AIs differ
·
Analyst track record
·
Who are the analysts?
·
Methods log
·
Experiments and rules fixed in advance
·
Corrections and replaced versions
·
What changed this week
·
Weighting simulator
·
Public API and widgets
·
System status
·
About
·
Compare two weeks
·
Timeline of AI developments
·
JJ HUB
·
Newsletter
·
Legal notice
·
Privacy
IT security watch
·
Daily digests archive
·
Story timeline
·
Compare days
·
Newly exploited vulnerabilities
·
Security methodology
·
Security sources
·
Security watch status
·
About the security watch
Researchers escape OpenAI Codex sandbox to run commands on host
50/100
Researchers escaped OpenAI's Codex sandbox, but the issue has been patched.