Meta is poised to fundamentally change how it evaluates the risks associated with new features and product updates across its platforms. According to internal documents obtained by NPR, the tech giant plans to automate up to 90% of its internal risk assessments using artificial intelligence, marking a significant departure from its decade-long reliance on human-led “privacy and integrity reviews.”
Traditionally, these human-led reviews have played a critical role in safeguarding user privacy, protecting minors, and preventing the spread of misinformation and harmful content. Under the new model, product teams will complete a structured questionnaire and receive instant, AI-generated feedback. The system will either greenlight the update or provide specific compliance requirements for the team to address through self-verification.
Meta asserts that the shift will streamline product development, enabling engineers to focus more on innovation while reserving human oversight for “novel and complex issues.” The company also emphasizes that the change reallocates human reviewer capacity to focus on higher-risk areas more likely to violate its policies.
However, internal concerns have emerged regarding the scope of AI’s role in reviewing sensitive areas, including AI safety, youth protection, and moderation of violent content. Some employees fear that this reduction in human oversight could increase the likelihood of real-world harm.
“Insofar as this process functionally means more stuff launching faster, with less rigorous scrutiny and opposition, it means you’re creating higher risks,” noted a former Meta executive. Another current employee warned that removing human evaluators may undermine the nuanced perspective necessary to foresee potential consequences.
Meta has acknowledged these concerns, stating that AI decisions are subject to internal audits. The company also confirmed that its European operations will maintain stricter oversight, as required by the EU’s Digital Services Act. Oversight for EU user data and products will remain under Meta’s European headquarters in Ireland.
This move is part of a broader AI transformation underway at Meta. CEO Mark Zuckerberg recently announced that AI agents will soon write most of the company’s code, including for its Llama models, and are already outperforming average developers in debugging and productivity. Meta is concurrently building internal AI tools to expedite research and product development.
The company’s decision follows a broader industry trend. Google CEO Sundar Pichai recently stated that AI now generates 30% of the company’s code, while OpenAI CEO Sam Altman suggested that figure is already 50% in some firms.
Still, the timing of Meta’s transition has raised scrutiny. The company’s move to AI-driven reviews comes shortly after the termination of its fact-checking initiative and a loosening of its hate speech policies. Critics argue that Meta may be dismantling long-standing safeguards in favor of fewer restrictions and accelerated product cycles—potentially compromising user safety.
As Meta embraces this next phase of AI integration, the balance between speed, innovation, and accountability will remain central to public and regulatory scrutiny.
