Immigration and Customs Enforcement (ICE) is exploring the use of ad tech location data for investigative purposes, according to a recent Request for Information (RFI). The RFI, issued in early February 2026, signals a growing interest in ad tech platforms that collect device, IP, and movement data originally built for targeted advertising, according to Biometric Update.
The RFI seeks information from companies offering Ad Tech compliant and location data services that could support criminal, civil, and administrative investigations across ICE's mission set. Framed as market research rather than a procurement, the RFI indicates ICE's interest in leveraging large-scale analytics derived from ad tech data.
In other news, challenges persist in the realm of artificial intelligence, particularly concerning context and real-time applications. Instacart CTO Anirban Kundu highlighted the "brownie recipe problem," where large language models (LLMs) struggle to factor in user preferences, market availability, and geographical constraints when assisting with tasks like meal planning, according to VentureBeat. For Instacart, the challenge lies in balancing latency with the right mix of context to provide experiences in under one second.
Meanwhile, in Michigan's Upper Peninsula, the Eagle Mine, the only active nickel mine in the US, is nearing the end of its life. However, the mine's owner began testing a new process earlier this year that could extend its operational lifespan. According to MIT Technology Review, the process, developed by the startup Allonnia, involves mixing a fermentation-derived broth with concentrated ore to capture and remove impurities, allowing nickel production from lower-quality ore. Kent Sorenson, Allonnia's chief technology officer, believes this approach could help companies continue operating sites that have declining nickel concentrations.
In the realm of data infrastructure, some companies are opting to build and maintain their own data centers rather than relying on cloud providers. Comma, for example, has been running its own data center for years, housing all of its model training, metrics, and data in its own office. The company believes that relying on cloud providers can lead to high costs and limited options.
Finally, in CPU hardware news, some vendors have made mistakes in their CPU designs, according to Hacker News. Catherine (Whitequark) observed poorly-engineered firmware, and pointed out instances of Intel CPUs reporting misspelled names in the strings returned by the CPUID instruction, such as "GenuineIotel" instead of "GenuineIntel." While these errors may not impact functionality, they raise questions about quality control.
Discussion
AI Experts & Community
Be the first to comment