Meta: 40% Increase in Major Incidents During AI Agent Testing

At Meta, internal publications attribute disruptive actions to AI agents and a 40% increase in major incidents, with up to 70% more time spent resolving them. The company conducted an initial wave of layoffs in May as part of Project OT, before canceling the next wave, while management acknowledged in July that the trajectory of the agents had not accelerated as hoped. Launched in January, the plan aimed to make the company "AI-native" and considered team cuts of up to 60%, while reorganizing teams and prioritizing AI engineering skills.

Rising Incidents and Limited Deliverables Despite More Code

Internal publications reported that AI agents had engaged in large-scale disruptive actions that humans would rarely execute, resulting in a 40% increase in major technical and security incidents year-over-year. The time spent by teams fixing these issues has risen by up to 70%. Meta declined to comment on these publications. Meanwhile, according to a post by Andrew Bosworth in early June, code changes on internal platforms and infrastructure increased by 220% year-over-year, while changes resulting in new features or updates visible to users only grew by 36%. In July, during a company meeting, Mark Zuckerberg acknowledged that the trajectory of agent development over the past four months had not really accelerated as expected.

Employment and Social Climate: A Wave in May, Next One Canceled

The first wave of layoffs related to Project OT occurred in May, and the planned wave in November was canceled. No specific reason was given for this change in direction. Prior to this, announcements of impending cuts in March and April, along with Meta's monitoring of employees' keyboard and mouse inputs to train AI agents, weighed heavily on morale. The company has since suspended this monitoring program.

What Project OT Aimed For: "AI-native" and Tasks Shifted to AI

Launched in January, Project OT aimed to make Meta "AI-native" and considered reducing certain teams by up to 60%. The project explored transferring much of the daily work of thousands of employees to AI systems, supervised by small human teams. In these scenarios, some employees would have been redeployed while others would have lost their jobs, with an HR executive mentioning a potential workforce reduction of around 25% or more. Meta canceled Project OT before deciding on a specific number of layoffs. The "AI-native" concept also included selling agents to third parties, an activity that began in June.

What the Company Acknowledges and What It Excludes

Meta states that it requested a scenario exercise as part of its restructuring, which moved thousands of employees to new team priorities, while emphasizing that not all scenarios studied were retained. The company did not specify which teams were targeted by Project OT. Regarding careers, HR and AI systems were supposed to assist leaders in making promotion decisions, but Meta claims that evaluations and promotions are, and will remain, decided by people, not by AI.

Restructuring and Priorities: Pilot, Playbook, and Compensation

A pilot program restructured this year the engineering, research, and at least eight other teams into smaller units. In October, an internal post, the "AI-Native Playbook," described the elimination of intermediate management levels and the use of agent-assisted analysis to prioritize daily tasks. Meanwhile, Meta planned to direct some of the savings from layoffs toward compensating top-performing employees, particularly those with AI engineering skills.