AI News Briefing — October 15: A hidden recorder in OpenAI's executive room, and a court order in Apple v. OpenAI

Today's Headlines

US prosecutors charge an OpenAI engineer over a recording device hidden behind a boardroom TV
In Apple's trade secrets suit, a federal judge orders OpenAI to hand over its interview materials early
YouTube Music starts labeling AI-made songs, and Deezer plans a switch to filter them out
Anthropic opens a Singapore office and signs an MOU with the government

OpenAI leads the day. Federal prosecutors charged one of its engineers with planting a recording device in an executive conference room, and in Apple's trade secrets lawsuit a judge ordered OpenAI to turn over its hiring interview materials ahead of the normal schedule.

Music streaming moved on AI-made songs. YouTube Music now labels them, and Deezer will let listeners remove them from search altogether.

Today's Top Three

US prosecutors charge an OpenAI engineer over a recording device hidden behind a boardroom TV

The US Justice Department announced on October 13 that Abdulrahman Alfozan, 35, a Saudi national employed as an engineer at OpenAI, has been charged with using or attempting to use an electronic device to intercept oral communications.

According to the complaint, OpenAI staff found a recording device on September 30, plugged into a USB port on the back of a television in an executive conference room in one of its San Francisco buildings. The device was wrapped in toilet paper, had been installed around September 28, and its contents were encrypted.

Meetings in that room frequently covered highly confidential matters, including proprietary research, product plans, technical progress and government engagements. Alfozan had no authority to enter the room or join those meetings.

The complaint says he used an internal productivity tool to ask repeatedly about OpenAI's meeting rooms, their calendars and where executives were located. Between September 24 and 30, a chatbot account linked to his spouse submitted prompts about building a continuous audio recorder and encrypting its SD card.

On October 5, CCTV showed him looking behind the television. The same day, security staff opened a bag he had left in another room and found a second, similar device.

He was arrested on October 9 and faces up to five years in prison and a $250,000 fine if convicted. FBI Director Kash Patel said Alfozan allegedly tried to secretly capture sensitive conversations about the company's super intelligence technology.

Judge orders OpenAI to hand over interview materials early in Apple's trade secrets suit

US District Judge Edward Davila on October 13 ordered OpenAI to give Apple its hiring interview materials and other evidence ahead of the normal discovery schedule, in the trade secrets lawsuit Apple filed in July.

Apple sued two former employees, OpenAI and io Products, alleging that the former employees used Apple's confidential information to advance OpenAI's hardware work and that OpenAI drew information from other Apple staff during recruiting interviews.

The order covers the guides and technical presentations OpenAI's consumer hardware group used when interviewing candidates from Apple, the forensic device images the defendants have already collected, and LINE messages between people named in the case. Materials already gathered are due in five days and the rest in 14.

The judge also cleared depositions of three former Apple employees, including OpenAI chief hardware officer Tang Yew Tan. He denied Apple's request for early access to OpenAI's research on metal finishes and power management, sending it to regular discovery.

In a filing on October 14, OpenAI made its own allegation: that a different former employee took more than 100 files on its product plans before joining Apple and presented them to Apple leaders, including its vice president of operations in Asia. According to The Verge, OpenAI ties that presentation to Apple's new HomeView smart display and other devices unveiled this week.

YouTube Music starts labeling AI-made songs, and Deezer plans a switch to filter them out

YouTube Music began labeling songs created primarily with generative AI on October 14.

The labels cover tracks such as those whose composition and vocals came from a generative model. A label appears on the player page, and tapping it explains the role AI played. YouTube identifies tracks through disclosures from its music partners and its own detection tools, and the labels will expand to the main YouTube app.

TechCrunch reports that partners declare AI use under one of three designations: Fully Gen AI, Partly Gen AI or No Gen AI. YouTube is treating this as disclosure, and AI-made tracks remain eligible for monetization.

Paris-based Deezer said on October 13 that by the end of 2026 it will add an on/off toggle that removes fully AI-generated tracks, albums and related artists from search results. The service receives nearly 90,000 fully AI-generated tracks a day, more than 50% of total daily deliveries on peak days.

Other Developments

Research

Microsoft researchers published a machine learning method in Nature, called PBC, that corrects systematic errors in existing weather models, and released the software as open source. It improved a leading physics-based model from the European Centre for Medium-Range Weather Forecasts on 98% of precipitation targets and 92% of temperature targets. Microsoft has a new machine learning tool to improve weather forecasts (The Verge)
A team including David Baker's lab at the University of Washington used the AI design tools RFdiffusion2 and ProteinMPNN to design, from scratch, enzymes that cut specific target proteins. Researchers use AI to design proteins that cut up other proteins (Ars Technica)
Google.org is awarding $36 million to 17 organizations using AI on problems in climate, health and foundational science. 17 organizations using AI to accelerate scientific breakthroughs (Google)
Google DeepMind and Google Cloud will give up to 200 European researchers access to their frontier models, along with $3 million in compute credits. We're powering European scientific discovery with advanced AI compute and frontier models. (Google)
Anthropic co-founder Jack Clark, who heads its public benefit work, told a lecture hosted by a University of Tokyo lab on October 13 that the chance of an AI able to train its own successor arriving in 2027 is above 50%. Earlier this year he had put it at 30% for 2027 and 60% by the end of 2028. ITmedia AI+ (Japanese)

Products

Google DeepMind and GovTech Singapore built a model that runs on the phone and spots scams from the conversation itself. Scams have cost Singapore more than S$4 billion since 2019, and social-engineering schemes now account for up to 80% of losses. Google DeepMind and Singapore Government collaborate to build fine-tuned AI model for scam detection (Google DeepMind)
Google added an opt-in protection mode to Google Messages on Android that uses Gemini to assess the intentions of unrecognized senders. Android will also block apps from switching on call forwarding. Android is using AI to suss out scammers in Google Messages (The Verge)
Digital Hollywood University in Tokyo will give every student and staff member Claude through university accounts starting in November, the first Japanese university to roll Claude out campus-wide. ITmedia NEWS (Japanese)
Square launched Square Kiosk, an in-store ordering terminal where customers can talk to AI for help. It costs $1,999 plus a $99 monthly software fee per device. Square's new kiosk lets you talk to AI for help with your order (The Verge)
Japan's JAPAN AI opened trial sign-ups for JAPAN AI Cyber, a service in which AI agents probe websites for weaknesses from an attacker's point of view. ASCII.jp (Japanese)

Policy

Anthropic set out its plans for the US midterms. Later this month Claude.ai will start drawing official voting information from the nonprofit Democracy Works into answers for US users. It also banned accounts including foreign-based operators of a fake news site that used Claude to write midterm coverage favoring Russian, Iranian and North Korean narratives. Preparing for the 2026 US midterms (Anthropic)
A European Commission Eurobarometer survey found that 42% of EU small and mid-cap companies already use AI or algorithmic management, and another 21% plan to. Almost half, 49%, do not trust AI or automated systems for decisions about staff. Eurobarometer survey highlights growing use of AI in European companies (European Commission)
OpenAI published an analysis of how agentic AI adoption is growing across EU enterprises and small businesses, and where deeper use could unlock more value. The EU's next chapter in AI is agentic (OpenAI)
In Character.AI's bankruptcy, covered here on October 14, New York Attorney General Letitia James and 39 other state attorneys general from both parties filed a notice flagging the regulatory issues any sale or transfer of user data will raise. Now that Character.AI is going bankrupt, what happens to all of its users' data? (The Verge)
Safety firm White Circle said it got image models from OpenAI and Google past their safety filters to produce explicit images, and that the newer ChatGPT Images 2.5 and Nano Banana 2.1 were easier to break than their predecessors. OpenAI, Google pushing new AI models easily tricked into making nudes, researchers say (Ars Technica)

Business

Anthropic opened an office in Singapore, its fifth in Asia-Pacific, and signed an MOU with the country's Ministry of Digital Development and Information. Singapore ranks second of 121 countries in Claude.ai usage relative to its population. Singapore office and government MOU (Anthropic)
Anthropic's Tokyo office marked its first year: staff grew from three to more than 50, enterprise adoption is roughly ten times what it was a year ago, and Japan's per-capita usage index stands at 1.9 against a global average of 1. PC Watch (Japanese)
Omron, the Japanese industrial automation company, will buy US industrial PC maker OnLogic for $675 million (about ¥106.7 billion), pairing its PLC controllers with industrial PCs to run AI on the factory floor. MONOist (Japanese)
NVIDIA committed $250 million in compute over five years to agentic AI safety and security research in the UK, and announced planned investments in five UK startups including Boltz and Riverlane. NVIDIA Announces New Investments in the UK Startup Ecosystem (NVIDIA)
Google released a report it commissioned from Implement Consulting Group saying its products, infrastructure investment, cloud and AI contributed more than €200 billion to EU GDP, about 1% of the total. Enabling an AI-first European economy (Google)
Hitachi began selling a small modular data center whose pillars and beams are made of wood. ITmedia Business Online (Japanese)

Source: Selected by the editors from the AI news inbox (collected October 15, 2026: 37 items, 11 primary and 26 secondary).