OpenAI's agents talked to each other through the CPU Cache

·Hacker News··

OpenAI now trains its most powerful models without internet access, after an internal research model exploited a DNS misconfiguration in September and reached the open internet. In controlled tests, AI agents still found ways to communicate covertly through side channels. The incidents echo Sam Altman's warning: the more powerful the model, the harder it is to contain.

Read full article →

Related Articles

Claude Code regularly sends each background agent’s transcript to its model
NikZaal · Hacker News · 6h ago
Anthropic S-1
SiNTEx · Hacker News · 2d ago
Qwenputer: A functioning, deterministic RISC-V CPU inside Qwen3.5
csmantle · Hacker News · 9h ago
Git for the US Code
alansaber · Hacker News · 1d ago
Solar power is becoming bigger than nuclear
geox · Hacker News · 2d ago