Meta will reportedly automate most of the risk assessment process
Internal Meta documents obtained exclusively by American public radio revealed that Meta plans to automate many of its product risk assessments. Before Meta changed its name, back when it was still Facebook in 2012, it settled a lawsuit with the Federal Trade Commission (FTC) over privacy settings. As part of the settlement, Meta was required to conduct a privacy review of its products and evaluate the risks of potential updates.
Until now, this privacy review of Meta products has been carried out by human evaluators. However, according to the internal documents, up to 90% of this process could be replaced by AI evaluation. The new AI privacy review system will require product teams to complete a survey about their work, which will then instantly identify any privacy risks and requirements that updates or features must meet before being released.
AI-based Approach for Faster Updates
The AI-based approach will enable Meta to update its products more quickly. Nevertheless, this shift also raises concerns about the potential risks of replacing humans with AI. A former Meta executive, speaking on condition of anonymity, expressed concerns that relying more on AI could lead to negative consequences that are not identified before causing problems.
In response to a request from TechCrunch, a Meta spokesperson highlighted that the company has invested over $8 billion in its privacy program. The spokesperson emphasized Meta's commitment to delivering innovative products while meeting regulatory obligations. The spokesperson also mentioned that as risks evolve and programs mature, Meta will enhance its processes to better identify risks, streamline decision-making, and improve the employee experience.
Meta leverages technology to bring more consistency and predictability to low-risk decisions. However, the company also values human expertise for the rigorous evaluation and oversight of new or complex issues, emphasizing the importance of human privacy review in addition to AI automation.