The Bug Wrapped in Toilet Paper: Silicon Valley's Spy Scare Comes From Inside

The Bug Wrapped in Toilet Paper: Silicon Valley's Spy Scare Comes From Inside

A sweep, a USB port, a roll of toilet paper

On Sept. 30, a technical surveillance countermeasures team sweeping an executive conference room at OpenAI's San Francisco office found a thumb-sized recorder plugged into the USB port on the back of a wall-mounted television and wrapped in toilet paper. According to the federal complaint, it had been installed around Sept. 28 and its contents were encrypted. The room could be booked only by executives and their assistants, and hosted leadership meetings, product planning and briefings with government officials.

The engineer who was supposed to be there

Prosecutors charged Abdulrahman Alfozan, 35, a Saudi national who joined OpenAI as an engineer in July 2025. On Oct. 5, security staff searching his bag found a second, near-identical device, and a covert camera installed behind the same television recorded him returning to check on the first, according to court filings. He was arrested on Oct. 9; OpenAI fired him that day.

What the charge does, and does not, say

The Justice Department charged him under 18 U.S.C. § 2511 with using or endeavoring to use an electronic device to intercept oral communications. The maximum penalty is five years in prison, a $250,000 fine and three years of supervised release. He has not been charged with espionage, and no foreign government has been named. A detention hearing is set for Oct. 19 before Magistrate Judge Alex G. Tse.

A one-way ticket and a venture firm

Prosecutors told a judge that before his arrest Alfozan had booked one-way tickets from San Francisco to Saudi Arabia for himself, his wife and a third family member. CNBC reported that he is also a partner at Raed Ventures, a Saudi venture firm whose website described him as an engineering lead at OpenAI. The Financial Times reported that when Saudi tech influencers visited San Francisco in September, he was the one who showed them around OpenAI's offices. His lawyer, William Pernik, said his client was charged with violating a recording statute, "not stealing trade secrets or working as a foreign agent," adding: "Where someone was born is not evidence of anything."

OpenAI's careful wording

OpenAI said it found "a sophisticated recording device" and used its own models to work out where it came from, identifying the employee who planted it. The company said it had "no indications" that he improperly accessed or exfiltrated model weights, source code or customer data, and that it had referred the matter to the FBI. The detail that landed hardest in the Valley: the lab's products were used to police the lab.

Why this one touched a nerve

Silicon Valley has spent years hardening its networks against hackers. A microphone behind a television walks past all of it. The Wall Street Journal reported that OpenAI already trains staff on "honey pot" approaches — one employee was approached by a stranger at a San Francisco playground who steered the conversation toward their work — and advises screen protectors on transit and wiped whiteboards. Jacob Coxon, a former OpenAI and Anthropic researcher, said he is 90 percent certain that Chinese spies are already inside America's leading labs. Hiring is leaky too: Pindrop says one in 47 applicants it reviewed this year showed signs of North Korean ties, a sevenfold increase.

The counterintelligence budget

Anthropic said in August that it had slowed data transfer in "sensitive environments" and protected some technology with physical security keys, while conceding that defending against sophisticated insiders and nation-state attackers "requires security investments beyond what we've currently achieved." OpenAI has hired Thomas Lind, formerly the AI policy lead at the White House Office of the National Cyber Director, to lead cyber and strategic risk.

Buying the expertise

The clearest signal is a deal that hasn't happened yet. The Information reported that OpenAI is in early talks to acquire Trail of Bits, the New York security firm founded in 2012 with about 40 employees, which has audited Uniswap, Aave and Worldcoin. No price has surfaced, and the talks could still collapse. The two already run "Patch the Planet," launched in June, which has flagged 1,917 issues across 67 codebases and merged 462 patches upstream. For a company building a cyber business to compete with Anthropic, buying a respected audit shop is faster than growing one.

What happens after Oct. 19

The detention hearing will decide whether Alfozan waits for trial in custody. Whatever the outcome, the operative assumption inside AI labs has flipped: the conference room is no longer assumed to be clean. As Jeffrey Ladish, a former Anthropic security official, put it, the threat "can no longer be dismissed as speculative. It has now actually happened at least once."