Espionage Comes For AI

The world may fret over AI’s potential to create novel security threats, but an older, more familiar danger has arrived inside the office of one of the industry’s biggest firms. A few days ago, OpenAI disclosed that its security team had found a recording device plugged into the USB port of a TV in a San Francisco conference room used for highly sensitive discussions. The bug had been planted by an employee, Abdulrahman Alfozan, who joined the company last year. Investigators later determined he intended to place additional recording devices in other restricted areas of the office.

Corporate espionage is not new. Firms have often spied on other firms. Even tech giants like Apple and Twitter (now known as X) have dealt with former employees who walked out with trade secrets or attempted to spy. What sets this case apart is that this is the first publicly recorded instance of someone being caught spying inside a prominent AI firm.

What has further piqued curiosity is Alfozan’s background. A Saudi national, he received a scholarship that allowed him to study computer science and engineering before moving to the US for further education. He later worked at Meta and at least one other startup before joining OpenAI. Public records and reporting have also linked him to a network of Saudi tech professionals in Silicon Valley who advised Saudi organisations, including the kingdom’s social insurance agency. He was also listed as a venture partner at a Riyadh-based firm that counts Prince Khalid bin Bader Al Saud among its partners.

None of these associations prove that Alfozan was acting on behalf of a foreign government. Yet, in an industry racing to develop technologies with profound national-security implications, these associations have raised eyebrows. Interestingly, back in 2019, two Saudi nationals working at Twitter had been accused of spying on the platform’s users, including dissidents of Saudi Arabia, on behalf of the Gulf state. And just last week, Jacob Coxon, the former Anthropic researcher whose warnings about the risks of AI had spooked much of the world, raised alarms about the possibility of spies inside the offices of leading AI firms. In a podcast, he claimed to be ‘90 percent’ certain that China has spies inside Anthropic and OpenAI.

Alfozan may not have been able to secure any information of value, as has been claimed, but the revelation that someone was caught attempting to do so, will alarm the industry. As AI capabilities grow more strategically valuable, more and more individuals, firms, and probably even countries, will look to steal information of value. This OpenAI discovery may be the first confirmed case of spying inside an AI firm, but it is unlikely to be the last.