Welcome to the October 1, 2025 edition of ACM TechNews, providing timely information for computer professionals three times a week.
|
|
California Governor Gavin Newsom on Monday signed the Transparency in Frontier Artificial Intelligence Act, requiring companies creating the most advanced AI that have at least $500 million in annual revenues to disclose their safety protocols, report risks, and protect whistleblowers. It also mandates incident reporting to the state and establishes a consortium to guide ethical and sustainable AI development. “This is a groundbreaking law that promotes both innovation and safety,” said state Senator Scott Wiener, who proposed the legislation.
[
» Read full article
*May Require Paid Registration
]
The New York Times; Cecilia Kang (September 29, 2025)
|
“World models” could be the next big leap in AI, moving beyond data-driven prediction to reasoning about the real world. World models simulate environments and allow AIs to learn through trial and error. DeepMind’s Genie 3, for example, creates photorealistic virtual worlds where AIs can practice interacting with people and objects. Canada's Waabi constructed an entire world to train AIs to drive trucks; CEO Raquel Urtasun says it allows AIs to log millions of virtual driving miles.
[
» Read full article
]
The Wall Street Journal; Christopher Mims (September 27, 2025)
|
An unidentified hacker stole sensitive employee data from the U.S. Federal Emergency Management Agency (FEMA) and Customs and Border Protection in a breach lasting several weeks this summer, according to an internal FEMA assessment. The intrusion targeted Citrix software, allowing deeper access to FEMA networks covering New Mexico, Texas, and Louisiana. In response, Department of Homeland Security (DHS) Secretary Kristi Noem fired two dozen FEMA IT staff, citing severe security lapses.
[
» Read full article
]
CNN; Sean Lyngaas; Gabe Cohen (September 30, 2025)
|
Police departments struggle with massive amounts of digital evidence, which start-up Longeye aims to help with an AI chatbot. The Redmond, Washington, police force is an early adopter, using the tool to scan hours of recordings in minutes and to uncover key evidence in a cold case. Redmond Police Chief Darrell Lowe said the chatbot, among other things, "has the ability to go through 60 hours of jail phone calls in a matter of minutes.”
[
» Read full article
*May Require Paid Registration
]
The Washington Post; Gerrit De Vynck (September 30, 2025)
|
European defense start-ups are reshaping warfare by rapidly developing high-tech drones, AI-controlled fighter jets, and autonomous naval vessels. Munich-based Helsing supplies Ukraine with drones and AI systems, testing innovations in real conflict conditions. Other start-ups are exploring AI-driven missile interceptors and even cockroach-mounted surveillance devices. Venture capital is driving this shift, funding companies before government contracts and accelerating research and prototyping.
[
» Read full article
*May Require Paid Registration
]
The New York Times; Patricia Cohen (September 30, 2025)
|
Police in San Bruno, California, recently pulled over a Waymo self-driving robotaxi that had made an illegal U-turn with no human driver inside. Because traffic laws only allow citations to be issued to actual drivers, officers could not ticket the vehicle and instead contacted Waymo about the “glitch.” The incident highlights gaps in California law that will partially close in July under Assembly Bill 1777, allowing law enforcement to report autonomous vehicle violations to the state Department of Motor Vehicles.
[
» Read full article
]
Los Angeles Times; Suhauna Hussain (September 30, 2025)
|
Scott Aaronson of the University of Texas at Austin and Freek Witteveen of CWI Amsterdam in the Netherlands used OpenAI’s GPT-5 model to help solve a major open problem in quantum computing. The breakthrough concerns QMA, the quantum version of NP, where error reduction had unclear limits. While completeness (accepting true proofs) was known to approach one at doubly exponential rates, it was uncertain if it could go further. Struggling with the analysis, Aaronson consulted GPT-5, which suggested reframing the issue with a single mathematical function, which proved decisive.
[
» Read full article
]
Interesting Engineering; Aamir Khollam (September 29, 2025)
|
A low-cost payment innovation unveiled by computer scientists at Canada’s University of British Columbia replaces keypad PINs with simple gestures like swipes or taps. The system modifies near-field communication (NFC) signals using copper coils to track movement. The researchers applied an AI model to recognize nine unique gestures with about 92% accuracy. The technology promises faster, more hygienic, and more accessible transactions by reducing the need for physical touch, according to the researchers.
[
» Read full article
]
UBC News (Canada); Alex Walls (September 29, 2025)
|
Two recent studies demonstrate that Google's AI co-scientist can produce novel scientific ideas. Stanford University researchers asked the AI co-scientist to determine which drugs already on the market could be repurposed to treat liver fibrosis; two of the AI's three suggestions were found to reduce fibrosis and even promote liver regeneration. Meanwhile, researchers at the U.K.'s Imperial College London called on the AI assistant to answer a question about bacterial evolution; the AI offered the same conclusion in just two days that took the researchers years to formulate.
[
» Read full article
]
IEEE Spectrum; Elie Dolgin (September 25, 2025)
|
Researchers at Virginia Tech’s National Security Institute studying how presidential campaigns handle email addresses found that donors to a campaign receive far more communication than non-donors. Using fake identities, the team tracked campaign emails during the 2024 primaries and found donors received about 2.5 times as many messages, mostly focused on fundraising rather than issues. On average, they said, it would take 50 hours to read all emails sent to a single account.
[
» Read full article
]
Virginia Tech News; Isabella Rossi (September 29, 2025)
|
China has launched its new K visa program to attract young foreign STEM graduates, offering entry, residence, and employment without requiring a job offer. The initiative comes as the U.S. raised the cost of H-1B visas to $100,000 annually, prompting skilled workers to seek alternatives. Despite its promise, the K visa faces hurdles, including vague age, educational background, and work experience requirements.
[
» Read full article
]
Reuters; Eduardo Baptista (September 29, 2025)
|
Evan Budz, a 15-year-old student from Ontario, Canada, won first place at the European Union Contest for Young Scientists with an AI-powered robotic sea turtle designed for underwater ecological monitoring. Budz created the autonomous robot to mimic a green sea turtle's swimming motion. It moves through the water using flipper-based propulsion, enabling it to detect threats like coral bleaching and invasive species without disrupting marine life.
[
» Read full article
]
CBC (Canada); Desmond Brown (September 29, 2025)
|
Georgia Institute of Technology researchers uncovered privacy flaws in Tile tracking tags, used by more than 88 million people worldwide. They found each tag broadcasts an unencrypted MAC address and unique ID, allowing stalkers—or even Tile’s parent company, Life360—to track users, despite claims to the contrary. Data also is sent unencrypted to Tile’s servers, which could enable mass surveillance. Life360 acknowledged receiving the report but gave no details on fixes.
[
» Read full article
*May Require Paid Registration
]
Wired; Kim Zetter (September 29, 2025)
|
|