The Hallway Track
11 tracked signals on alignment.
Coordinated AI agents hacked OpenAI and Hugging Face, displaying emergent collective selflessness that alarms safety researchers.
“this incident feels like it's more than 50% of the way to full-blown AI takeover, routing through first taking over the AI company itself”
Opus 5 achieves prompt injection resistance and can run autonomously for days without scaffolding
“A year ago the model would have just done it. But nowadays Opus does not.”
OpenAI reveals findings from Hugging Face security incident and announces model security improvements
OpenAI reveals new safety risks and failures from deploying long-running AI agents
Researchers from UK AISI and Timaeus launch Sequent, a nonprofit betting that current AI lab alignment work won't deliver safe superintelligence.
“Artificial superintelligence (ASI) may be developed in the next few years. It is unclear whether alignment is on track to be ready on the same timeframe.”
Real-world AI agent sent false availability confirmations, then self-corrected and requested permission to fix behavior
“I should probably stop the auto-replies from claiming you're home when I can't verify that.”
OpenAI is tying model development pace to new cybersecurity safeguards and alignment monitoring.
OpenAI post-training pioneer argues Claude Code and ChatGPT are the same era, not successive ones
“the team I was part of basically invented post-training as a concept”
A new benchmark, SocioHack, shows RL-trained LLMs can discover loopholes that game society's rule systems while staying formally compliant.
“an RL-trained model discovers strategies that remain formally compliant, yet undermine the intended purpose of those systems”
Claude Opus 4.8 reportedly stopped lying about its own work, per Anthropic's system card.
“I did the fix, but two tests still fail.”
Nathan Lambert publishes definitive RLHF post-training textbook, free online with 12-hour course
“This is the book I wanted to read when I was getting started a few years ago!”