The development of an artificial intelligence system (model) that is capable of improving itself without human intervention and developing its next models is accelerating, experts are calling it autonomous recursive self-improvement.

According to a report by The News International, Anthropic revealed that its cloud system is taking about 26 percent of the research and development work on its new models under human supervision, but the artificial intelligence is not yet fully autonomous.

OpenAI has set a goal of fully automated artificial intelligence research by March 2028, but the organization admits that how to keep such automated improvements safe and under human control is not yet clear.

SpaceX CEO Elon Musk says human supervision at Grok is gradually decreasing and fully automated improvements could be possible by the end of 2026 or 2027.

Microsoft, on the other hand, is pushing for human supervision and the development of artificial intelligence within clear boundaries.

According to experts included in the report, there are fears that artificial intelligence could be difficult to control if it starts to improve its capabilities faster than humans.

OpenAI also cautions that research on safety and human compatibility is not guaranteed to keep pace with the growth of artificial intelligence capabilities.