Google's AI struggles with police data requests management

Google is struggling to use artificial intelligence (AI) to manage the large number of requests from police for user data. In the first half of 2024, the company received 236,000 requests, overwhelming its Legal Investigations Support (LIS) team. To cope, Google attempted to develop AI tools that could process these requests more quickly. However, current and former LIS employees reported that these AI tools have not worked as intended. Although the AI was designed to understand and process court orders, it has failed to perform the necessary tasks effectively. Sources say the backlog of requests is still in the thousands. Moreover, Google recently laid off 10 engineers who were working on the AI project. This has raised doubts about the future of the initiative. Some team members noted that the AI has not been deployed and its performance has actually made the situation worse, as human reviewers are still needed to check and redo much of the work done by AI. Experts have criticized the idea of using AI in legal processes. Cooper Quintin from the Electronic Frontier Foundation emphasized that AI can make mistakes, potentially leading to serious issues. He expressed concern that AI could struggle to differentiate between valid requests and fraudulent ones, especially given the increase in hackers posing as law enforcement. Google has not commented on the layoffs or the effectiveness of its AI solutions. The company states it is continuously working to improve efficiency without changing how it handles law enforcement requests. However, experts warn that relying on AI could complicate the existing problems with fake orders.