Technology

AI labs report progress toward autonomous self-improvement as oversight questions remain

Developers are expanding model-driven development while retaining human oversight, according to ABC News.

Updated Sep 19, 2026 · 12:19 UTC1 reports

The short version

Leading technology developers are giving artificial intelligence systems expanding roles in designing and improving subsequent models.[ABC News]
Anthropic reported its Claude system directs 26% of model development under human oversight, while OpenAI unveiled an automated research intern.[ABC News]
Elon Musk stated xAI aims for near-full automation by the end of the year or by 2027, though models are not yet fully autonomous.[ABC News]
Companies disagree on definitions of recursive self-improvement, and Anthropic indicated it would pause development only if global rivals verifiably do the same.[ABC News]

Anthropic reported that its Claude system leads 26% of company model research and development under human supervision.[ABC News]
OpenAI introduced an automated research intern to execute defined tasks and aims to develop an automated researcher by March 2028.[ABC News]
Elon Musk stated that human involvement in xAI's Grok models is decreasing, with full automation targeted between the end of this year and no later than 2027.[ABC News]
Anthropic stated it would pause or slow development only if global competitors verifiably did so as well.[ABC News]

Artificial intelligence developers disagree on whether recursive self-improvement requires full autonomy or simply feedback from models during training.[ABC News]
OpenAI warned that rapid recursive self-improvement might not be an outcome labs should pursue.[ABC News]

Outlet counts describe coverage, not independent confirmation. Reports may share a wire service or original source.