Mark Zuckerberg, chief executive officer (CEO) of Meta Platforms, said on the 2nd (local time) that development of artificial intelligence (AI) agents is progressing more slowly than expected.
Zuckerberg said at a company town hall meeting that day that during January to February, when the company was drawing up a restructuring plan, top executives were concerned that the company was not moving fast enough to adapt to change. He added that at the time, leadership was "very optimistic" about models such as Claude Code from Anthropic, a rival.
Looking back, Zuckerberg said, "At least over the past four months, it seems AI agent development has not accelerated in the way we expected," noting that the company's bet on the new structure has yet to bear fruit. However, he said Meta expects to start seeing greater benefits from its AI investments within the next three to six months.
Also that day, Andrew Bosworth, Meta's chief technology officer (CTO), said that a review of the recent data security incident found that employee data was not included in AI training. Last month, Meta temporarily paused the program—which tracks employees' mouse movements and digital activity for AI training—while investigating the exposure of sensitive data.
Bosworth said that even if the company restarts the program, it will proceed on an "opt-in (consent-based participation)" basis. When Meta first installed the program on U.S. employees' computers in April, Bosworth said there was no way to opt out (refuse).