AI news — March 16, 2026
The 10 top stories PulseAugur surfaced that day, ranked by signal across labs, papers, and developer communities.
-
FSF seeks user freedom in Anthropic's LLM training data copyright settlement
The Free Software Foundation (FSF) has commented on the settlement in the Bartz v. Anthropic copyright infringement lawsuit. This class action suit alleges Anthropic used copyrighted materials from datasets like Library Genesis to train its large language models. While a court i…
-
AI falls short in cancer fight, experts say; deeper AI needed
AI has not yet delivered on its promise to significantly impact cancer treatment, despite hopes that advanced algorithms could revolutionize drug discovery and development. Critics argue that the challenges in medicine are often rooted in economic incentives and a complex unders…
-
xAI releases Grok's Text to Speech API with natural, low-latency voices
xAI has launched its Grok Text to Speech API, enabling developers to integrate natural and expressive voice capabilities into their applications. The API supports low-latency streaming and is multilingual, offering over 20 languages. It is designed for easy integration, requirin…
-
Google tests LLMs as expert partners for superconductivity research
Google researchers have published a paper evaluating the ability of large language models to act as expert research partners in the field of high-temperature superconductivity. The study involved posing complex questions about this area of physics to six different LLMs and havin…
-
Tower Semiconductor and Scintil Photonics debut single-chip laser for AI data center optics
Tower Semiconductor and Scintil Photonics have developed the first single-chip DWDM light engine designed for AI data centers. This technology uses dense wavelength division multiplexing to send multiple optical signals over a single fiber, which can significantly reduce power c…
-
Open models lag behind closed AI labs, but maintain significant influence
The AI landscape is seeing a significant push towards open models, driven by companies hoping to gain influence and mindshare, even if direct monetary returns are unclear. While open models have consistently lagged behind top closed models by 6-18 months, their rapid progress on…
-
AI agents can post-train LLMs, but humans still outperform them
A new benchmark called PostTrainBench has been developed to evaluate the ability of AI agents to autonomously refine existing language models for new tasks. While current AI agents can improve model performance, they still significantly underperform human capabilities in this ar…
-
Anthropic sues Trump admin, xAI restarts, Iran AI fakes cause chaos
Anthropic has initiated legal action against the Trump administration concerning AI technology within the Pentagon. This lawsuit is supported by an amicus brief filed by employees from OpenAI and Google. Concurrently, an internal Pentagon directive mandates the removal of Anthro…
-
Taiwan's ITRI institute drove chip dominance via successful industrial policy
The Industrial Technology Research Institute (ITRI) was established in Taiwan in 1973 with the goal of rapidly advancing the nation's technological capabilities, particularly in electronics. This government-backed initiative successfully transformed Taiwan from a low-cost manufa…
-
OpenAI explains why its security tool bypasses traditional SAST analysis
OpenAI's Codex Security tool takes a novel approach to code analysis by focusing on the repository's architecture and intended behavior rather than solely relying on traditional static application security testing (SAST) reports. The system is designed to validate findings again…