Meta is reportedly planning to transition a significant portion of its product and feature risk assessments to artificial intelligence (AI). According to sources, the Menlo Park-based social media company is exploring the possibility of allowing AI to manage the approvals of its features and updates, a task that has predominantly been performed by human evaluators until now. This shift is expected to influence the implementation of new algorithms, the introduction of safety features, and the mechanisms for sharing content across its various platforms. The company anticipates that this change will accelerate the rollout of new updates and products.
An NPR report indicates that Meta aims to automate up to 90 percent of its internal risk assessments. The publication claims to have obtained internal documents that outline this strategic shift.
Currently, any new features or updates for platforms such as Instagram, WhatsApp, Facebook, or Threads must be vetted by a team of human experts. These experts evaluate various aspects, including potential user impact, privacy violations, and risks to minors. Known as privacy and integrity reviews, these evaluations also consider the likelihood of feature-related misinformation or harmful content.
Under the proposed AI system, product teams can expect to receive an “instant decision” after completing a questionnaire related to a new feature. The AI will either approve the feature or generate a list of criteria that must be addressed before proceeding with the project. The product team will then need to ensure that these requirements are satisfied prior to the feature’s launch.
The report suggests that Meta believes the shift to AI will significantly enhance the pace at which features and app updates are released, allowing product teams to operate more efficiently. Nonetheless, some current and former employees at Meta have expressed concerns about whether this speed will compromise rigorous scrutiny.
In response to the report, Meta stated that human reviewers would still be engaged for “novel and complex issues,” asserting that AI would only oversee low-risk decisions. However, internal documents suggest that the transition may also empower AI to handle critical areas involving AI safety, youth risk, and integrity, including sensitive topics such as violent content and misinformation.
An anonymous Meta employee familiar with the risk assessment processes told NPR that the automation initiative began in April and has progressed through May. “I think it’s fairly irresponsible given the intentions behind our existence. We provide the human perspective on how things can go wrong,” the employee remarked.
Earlier this week, Meta published its Integrity Reports for the first quarter of 2025. In this report, the company noted, “We are beginning to see LLMs operating beyond that of human performance for select policy areas.”
The social media giant also mentioned that it has commenced utilizing AI models to filter content from review queues, particularly in cases where it can be “highly confident” that the content does not breach its policies. By doing so, Meta aims to optimize the capacity of human reviewers, allowing them to focus their expertise on content that is more likely to violate guidelines.
