OpenAI Engineer Charged Over a Hidden Recorder, Grok Orders Musk's Lego, and Fake Bylines Place 100 AI-Written Articles
Title: OpenAI Engineer Charged Over a Hidden Recorder, Grok Orders Musk's Lego, and Fake Bylines Place 100 AI-Written Articles
Date: 16 October 2026
Type: Blog
Author: SAASiQ (contact@saasiq.ai)
Word count: 1216 words
Reading time: 5 min
Published: 16-10-2026
An OpenAI engineer appeared in federal court in San Francisco on 13 October, charged over a recording device hidden in an executive conference room. Prosecutors say he used a chatbot to research how to build it. OpenAI says it used its own models to identify him. Elon Musk posted a screenshot of a Lego order placed for him by Grok on 11 October. The screenshot showed the order number. A YouTuber then asked Lego's AI customer-service assistant to cancel the order. OpenAI said on 8 October that an Iranian operation had used ChatGPT and seven invented writers to place about 100 articles in real publications. President Trump posted an AI-generated video of himself with a tiger on 13 October, three days after a man was killed at a wildlife park in South Yorkshire.
A recorder behind the television
Abdulrahman Alfozan, a 35-year-old engineer at OpenAI, was arrested on Friday 9 October. He made his first appearance in federal court on Tuesday 13 October and remains in custody. He is charged with using or attempting to use a device to intercept oral communications. The maximum penalty is five years in prison and a $250,000 fine.
Prosecutors say Alfozan plugged a small audio recorder into a USB port on the back of a television in an executive conference room at OpenAI's San Francisco office, on or about 28 September. Only executives and their assistants could book the room. He had no access to it. OpenAI staff found the device more than a week before the arrest.
According to the FBI's affidavit, a hidden camera recorded Alfozan checking the device. OpenAI's security staff also reported finding a second listening device in his bag. Reports say security footage showed him near the conference room on about a dozen occasions in September. Whether the device recorded any conversation has not been reported.
The complaint says that from 24 September Alfozan used a chatbot account belonging to his wife to ask how to build a continuous audio recorder, how to encrypt its memory card and how to add live listening. OpenAI told Fortune the chatbot was ChatGPT. Some reports say the complaint itself does not name the chatbot. His wife is not charged.
OpenAI says it used its own AI models to find the person who planted the device and to trace where it came from. It restricted Alfozan's access, put him on leave and then ended his employment on 9 October. "At this time, we have no indications of this individual inappropriately accessing or exfiltrating model weights, source code, or customer data," the company said. The FBI says its investigation continues.
Grok buys the Lego
On Sunday 11 October Elon Musk posted on X that he had asked Grok, the chatbot made by one of his companies, to choose and order Lego Star Wars sets for young children. He shared screenshots to show it had worked. They showed five sets, a total of $241.34 charged to an American Express card, and the order confirmation page.
The screenshots also showed a shipping address, a billing address, an email address, the order number and the last four digits of the card, according to several reports. Users on X said Musk had doxxed himself.
The YouTuber known as Enderman then posted screenshots of a conversation with Lego's AI customer-service assistant. In them, he appeared to give Musk's order number and ask for the order to be cancelled. "Just canceled your order, don't share your info online guys," he wrote. Another user posted screenshots that appeared to show the same Lego assistant reinstating the order and adding 400 more of the same product. None of the screenshots has been independently verified.
Musk replied that the cancellation had not gone through. He also said the address was not his home. "I don't live there. That's where my security team receives mail," he wrote. "That address has already been doxxed multiple times." Lego declined to confirm whether any of the activity affected the order.
Two AI agents were involved. Grok placed the order for Musk. Lego's assistant was then asked to cancel it by someone else, who gave the order number from Musk's post.
Seven writers who do not exist
OpenAI published a report on covert influence operations on Thursday 8 October. It covered two networks it had banned from ChatGPT. One, which OpenAI calls Dark Clark, was run from Russia and aimed at Latin America. The other, Bogus Bylines, was run from Iran.
Bogus Bylines used seven invented writers. OpenAI named them as Ervin B. Hoskins, Noah Lamington, Sophia Gonzalez, Michael Harrison, Ericka Feusier, Jenny Williams and Alice Johnson. The operators wrote prompts in Persian. They asked ChatGPT to improve English-language articles about the conflict between the US and Iran, to adapt them to each website's submission guidelines and to draft pitches to editors.
OpenAI says about 100 articles appeared in roughly a dozen outlets. The Washington Post, reporting on 9 October, counted at least 20 outlets. The earliest article OpenAI found dates from July 2025. OpenAI says the number of articles rose sharply after the fighting between the US and Iran began. It said the operation appeared to be commercial, and that it could not identify who was behind it.
One of the articles ran in the Port St. Joe Star, a weekly newspaper in Florida, on 22 September. It argued about the cost of American power in the conflict with Iran, under the Hoskins byline. It reached the paper through PeaceVoice, a syndication service the paper already used. The publisher told the Post he had not checked the writer's identity because he trusted that source. OpenAI's image-checking tool found that the author photo was AI-generated.
Op-eds under the Hoskins byline also appeared on Daily Kos and Middle East Monitor. Both told the Post they had removed the articles after hearing from OpenAI. PeaceVoice has tightened its policy on AI, the Post reported.
The tiger videos
A man was killed by a tiger at Yorkshire Wildlife Park near Doncaster on Saturday 10 October, after entering its enclosure. On Tuesday 13 October President Trump shared footage of the attack on Truth Social, with a caption warning people not to mess with nature.
Hours later he posted an AI-generated video of himself. It runs about 23 seconds and has no caption. It shows Trump wrestling a tiger and then befriending it, to the song "The Lion Sleeps Tonight". The fact-checking site Snopes confirmed that Trump had posted it. The fact-checkers at Lead Stories examined a separate montage of five animal-attack clips that Trump posted the same day. They found that three of the five were AI-generated.
The man's family asked people to stop sharing the footage. Their statement said the "deeply offensive media content" was causing them severe distress at an exceptionally vulnerable time. Oliver Coppard, the Mayor of South Yorkshire, said Trump had used AI to mock the death and the family's pain "for little more than a cheap laugh on social media". The broadcaster Piers Morgan called the post "appallingly tasteless" and asked Trump to delete it.
What happens next
Abdulrahman Alfozan is due back in federal court on 19 October. Lego has not said whether Musk's order was cancelled. OpenAI has banned the ChatGPT accounts behind Bogus Bylines and has not named the operator.
SAASiQ - Intelligent Solutions for SaaS ©
