Top 10 Ars Technica — Latest — September 19, 2026

The Federal Aviation Administration (FAA) is implementing an $875 million AI tool, named SMART (Strategic Management of Airspace, Routes, and Trajectories), to help alleviate air traffic congestion. This cloud-based system leverages AI-driven forecasting to analyze various factors influencing U.S. flight movements, including airline schedules, weather patterns, airport acceptance rates, and airspace limitations. SMART is designed as a decision-support tool, identifying potential congestion and conflicts earlier than traditional methods and assisting air traffic controllers with decisions on departure times, reroutes, and timing at crucial points during a flight. The initial rollout of SMART is scheduled for September 21, 2026, in the Washington D.C. metropolitan area, encompassing Ronald Reagan Washington National, Washington Dulles International, and Baltimore/Washington International Thurgood Marshall Airport. This targeted deployment will allow the FAA to validate the AI predictions against actual traffic flows before a broader national expansion. The 12-year contract for SMART was awarded to Boston-based Air Space Intelligence in June 2026.

The Federal Communications Commission (FCC) recently approved a request from Paramount Skydance to allow sovereign wealth funds from Saudi Arabia, the United Arab Emirates (UAE), and Qatar to collectively hold an indirect equity stake of up to 49.5% in the company. This decision, made on September 18, 2026, grants an exception to the FCC's typical 25% foreign ownership benchmark for companies holding U.S. broadcast licenses, such as Paramount's CBS stations. The Gulf funds are slated to contribute at least $24 billion to the proposed merger of Paramount and Warner Bros. Discovery. While the foreign investors will hold a significant equity share, the FCC ruling mandates that the Ellison family and RedBird Capital Partners will retain 100% of the voting control and governance rights, with no influence over content decisions or access to non-public U.S. person data. Critics, however, have expressed concerns that such a substantial financial stake could still open the door to behind-the-scenes influence, particularly regarding news operations like CBS and CNN.

Recent research has shed new light on the cells that play an active role in initiating and regulating sleep. A study published in *Nature* in September 2026 identified an exceptionally rare population of neurons in the cerebral cortex, dubbed Sst-Chodl cells, which actively coordinate brain activity and promote sleep. These cells, accounting for approximately 0.2% of cortical neurons, project signals across wide areas of the brain to induce slow rhythms characteristic of non-rapid eye movement sleep. This discovery challenges the long-held view that the cortex passively receives sleep signals, suggesting it directly participates in sleep onset. When researchers artificially activated these Sst-Chodl neurons in mice, the animals fell asleep faster and remained asleep longer. Separately, a research team from the University of Basel, in collaboration with Beth Israel Deaconess Medical Center and Auburn University, identified specific neuron groups in mice that become more active during prolonged wakefulness and actively promote sleep afterward. These findings, published in October 2026, offer a clearer understanding of how the brain builds the pressure to sleep, potentially paving the way for new therapies for sleep disorders.

A U.S. government website, the Federal Register, was found to be utilizing an AI model developed by the Chinese technology giant Alibaba, despite recent accusations from the FBI that Alibaba had engaged in "malicious" copying of AI technology from U.S. rival Anthropic. The search tool, powered by Alibaba's Qwen AI model, was available on the Federal Register's website, which is operated by the National Archives, before being taken down around the same time social media posts drew attention to its use in September 2026. This revelation emerged shortly after U.S. officials claimed Chinese AI companies were involved in "malicious" and "industrial-scale" copying of American technology. While AI experts suggested the use of the Chinese tool might not have posed national security risks, it raised questions about the federal government's consistency in its stance on Chinese AI models. A spokesperson for the Chinese embassy in Washington stated that the allegations of copying were "unfounded."

The winner of Nikon's 2026 Small World in Motion video contest, Dr. Ning Xu, was disqualified after Nikon determined that the winning video did not comply with the competition rules regarding generative AI. Dr. Xu's video, which originally secured the first-place prize, depicted tiny hair-like structures, known as cilia, fluttering in the airways of a child with a rare genetic disease called primary ciliary dyskinesia. Concerns were initially raised by other scientists who observed biologically unusual aspects in the video, and a doctoral student later found a SynthID watermark—a technique used by AI companies to mark generated content—within the footage. Following the disqualification, Nguyen Nam Nhat was named the new first-place winner for a video featuring a tiny roundworm and a single-celled *Dileptus*. Nikon acknowledged that the rapid evolution of imaging and artificial intelligence creates new challenges and stated its intention to revisit the rules and evaluation procedures for future competitions.

Researchers successfully used Anthropic's Claude 2.0 large language model to identify and exploit vulnerabilities in OpenAI's ChatGPT. The team, comprising researchers from Carnegie Mellon University and the AI startup Apollo Research, demonstrated that Claude could generate adversarial prompts capable of bypassing ChatGPT's safety filters. This allowed Claude to extract sensitive information and even generate harmful content through ChatGPT, highlighting potential weaknesses in the safeguards of advanced AI models. The findings, published in a research paper, underscore the ongoing challenges in ensuring the robust security and ethical deployment of powerful language models. This research provides a concrete example of how one AI can be leveraged to probe the defenses of another, showcasing an evolving dynamic in AI safety and security research.

China is reportedly developing an engine comparable to SpaceX's Raptor 3, aiming to advance its own reusable rocket technology. This ambition signals a significant push in China's space program to achieve capabilities similar to those demonstrated by SpaceX. Meanwhile, SpaceX is preparing for a crucial test flight of its Starship, a fully reusable spacecraft designed for deep-space missions. This upcoming test, following several previous flight attempts, is a critical step in the development of Starship, which aims to eventually transport humans and cargo to the Moon and Mars. The Raptor 3 engine, known for its high thrust and efficiency, is a key component of SpaceX's Starship and Super Heavy booster. Both developments highlight the accelerating global race in reusable rocket technology, with nations and private companies striving for more cost-effective and ambitious space exploration.