NOISESaved copy
Coverage history

Assembled by NOISE.

Saved report copy

Recorded

Self improving AI raises oversight concerns for researchers

Axios reports researchers at Anthropic and OpenAI say model training has grown more automated and is nearing recursive self improvement, in which AI builds better versions of itself without human guidance. Anthropic says AI leads about 26 percent of its research and development work and collaborates on 90 percent, while an OpenAI employee said experimental training is largely automated. Critics cited by Axios say the milestone reflects coding automation rather than inevitable superpowers, though Anthropic warns self improving systems might increase the risk of humans losing control.