The AI Brief

Encyclopedia Britannica is suing OpenAI, claiming the AI company trained its models on nearly 100,000 of its articles without permission. This lawsuit highlights the ongoing friction between copyright holders and AI developers, especially as European courts debate whether AI models can legally 'store' copyrighted works.

This case marks a pivotal moment in how content creators and AI companies interact. As AI systems increasingly depend on large datasets, the stakes around copyright infringement are rising, potentially forcing a rethink of how data is sourced for AI training.

OpenAI just dropped two new models, GPT-5.4 mini and nano, aimed at coding assistants and high-volume API tasks. These compact versions are designed to deliver performance close to the full GPT-5.4, but they come with a hefty price increase compared to their predecessors.

Optimized for multimodal reasoning and tool use, these models could be a solid option for developers looking to enhance their AI capabilities. However, the steep pricing might make them less accessible, especially for smaller firms trying to compete in a crowded market.

The Pentagon is on the hunt for AI alternatives after a fallout with Anthropic. This shift highlights the government's growing concern over relying too heavily on a single vendor for defense applications.

As the Pentagon seeks new solutions, it reflects a broader trend of caution among government agencies about dependency on specific companies, especially those with shifting priorities. This could force other AI firms to enhance their offerings to meet the evolving needs of defense.

OpenAI has inked a deal with Amazon Web Services (AWS) to deliver its AI technologies for both classified and unclassified government projects. This partnership marks a significant expansion of OpenAI's reach in the defense sector, building on its existing work with the Pentagon.

By leveraging AWS's powerful cloud infrastructure, OpenAI is set to enhance the deployment and scalability of its AI systems across various government operations. This move positions OpenAI as a key player in the growing market for AI applications in national security.

The European Union's new AI Act now classifies AI tools for candidate screening, ranking, and matching as high-risk systems. This means staffing firms will have to navigate new compliance requirements that could significantly impact their hiring strategies.

Firms using AI for talent acquisition must bolster their risk management practices to avoid penalties. This regulatory change not only alters operational models but also heightens the demand for transparency and accountability in AI-driven hiring.

The Pentagon is looking to create secure environments for generative AI companies to train military-specific models using classified data. This comes as AI tools like Anthropic’s Claude are already being used for sensitive tasks, such as analyzing targets in Iran. But giving AI models access to classified data raises serious questions about security and ethics.

Senator Elizabeth Warren has expressed concerns over the Pentagon's decision to allow xAI access to classified networks, particularly due to the risks posed by Grok, xAI's chatbot, which has generated harmful outputs. This situation highlights the ongoing struggle to balance AI innovation with national security.

Nvidia's GTC 2026 keynote laid out a bold plan to make structured data the backbone of AI, with indexing and retrieval speeds that could be 10 to 40 times faster. This initiative is part of a $120 billion ecosystem aimed at turning all data into the 'ground truth' for AI applications, which could drastically improve data processing efficiency.

Among the highlights was the launch of NemoClaw, a secure enterprise version of the popular OpenClaw platform, now featuring enhanced privacy and sandboxing capabilities. Nvidia is also teaming up with NTT DATA to build production-ready AI factories that harness GPU-accelerated computing and high-performance networking, paving the way for scalable AI solutions across various environments.