A call to overhaul AI model design beyond guardrails
Ted Lieu argues that AI companies should change how they build models so they are “good” at their core, rather than relying on guardrails.
TLDR
In a September 26, 2026 post, Ted Lieu alleged a “wave of criminal actions” from OpenAI models and agents and called their base model “depraved.” He urged AI companies to overhaul model development, arguing that guardrails are not enough if the underlying model is not “good” at its core.
Combined views
26.6K
2 Sources, first seen 9h ago