ABOUT 1 MONTH AGO • 12 MIN READ

The Bug Behind the TV - Issue #27 - Insider Threat Weekly

profile

Insider Threat Weekly

*make sure to check your spam folder for the signup acceptance email.

Desi has mentally checked out this week. Please check next week.


Headline: The Bug Behind the TV

  • An OpenAI engineer has been charged with trying to use a hidden device to record conversations in an executive conference room in one of the company's San Francisco buildings. Abdulrahman Alfozan, 35, joined OpenAI in July 2025. He was arrested on 9 October and is in custody.
  • OpenAI's security team found the device on 30 September. It was a thumb-sized circuit board, wrapped in toilet paper and plugged into a USB port behind a TV. Its memory card appears to have been encrypted. A second, similar device turned up in a bag he left in a printer room.
  • According to the complaint, he used a company productivity agent to ask about meeting rooms, room calendars and where leadership met. A chatbot account linked to his spouse allegedly asked how to build, hide and encrypt a continuous audio recorder, and how to build a second and third unit. OpenAI says it used its own AI models to help trace the device.
  • OpenAI says it has no sign he took model weights, source code or customer data. He faces up to five years in prison if convicted. His detention hearing is on 19 October. His lawyer did not respond to The New York Times.

This is an interesting case, and we'll keep covering it in the newsletter, especially once more information comes out in court. At the moment it's only a complaint against him. OpenAI and the FBI haven't made any foreign government links, and no motive has been put out. Cases like this are super interesting because there could be an underlying motive tied to what OpenAI is working on. Only time will tell once it goes to court.

Source: Criminal complaint, United States v. Alfozan, No. 3:26-mj-71329 (N.D. Cal.), copy hosted by Ars Technica Source: Saudi National Arrested for Installing Surreptitious Audio Recording Device at OpenAI (US Department of Justice) Source: OpenAI engineer arrested over alleged bugging of executive conference room (SF Standard)​


Insider Scoop: The Barometer Claude Coded

  • The Australian Institute of Insider Threats (AIIT) rated the global insider threat environment 68.3 out of 100, or "High", in its first Global Insider Threat Barometer. The score is built from 409 news stories AIIT chose between January and June 2026.
  • Insider Threat Weekly sent AIIT questions on the method. The first edition had two different headline scores, two sets of validation thresholds, and percentages ten times too high. AIIT fixed all three in a revised report on 14 October.
  • Founder Boaz Fischer told Insider Threat Weekly that he picked the stories himself and an AI sorted them: "As each record was collected, Claude AI assisted with analysing and classifying the information against the established variables and updating the intelligence spreadsheet."
  • "We did not conduct a formal inter-rater reliability assessment," Fischer said. There was no peer review or sensitivity testing, and the 409 records are not public.

This is an interesting way to look at insider threat incidents. Once the database is released to members, it'll be interesting to see which cases they looked at, and whether they publish the framework for how they assessed them. At the moment there are some numbers there, and it's a bit of a vibe check around putting numbers to things. It would have been good to just review the cases, bucket the elements of each one and report on that, rather than coming up with a score now. Putting out a score now makes it harder for readers to shift their view if and when the framework changes in later years.

Source: 2026 Global Insider Threat Barometer Report (AIIT)​

Disclosure: I work for DTEX. This review was independent and DTEX was not involved.


Crime Roundup

The CEO and the chairman who looted their own bank.

  • Tomás Niembro Concha, former CEO of Nodus International Bank in Puerto Rico, was sentenced on 21 September to 112 months in prison. He took at least $24.9 million from the bank and conspired to evade US sanctions on Venezuela.
  • Board chairman and co-owner Juan Francisco Ramirez was sentenced on 7 October to 62 months. He took more than $23.6 million.
  • The pair had the bank put $11 million into a Miami lender, which lent it back to them. They also got the board to buy about $25.3 million of promissory notes from a finance company they jointly owned. The money went on mortgages, credit cards and personal investments.
  • They hid it from the rest of the board and the regulator. The bank failed in 2023.

Source: Former bank CEO sentenced to over 9 years in prison for multimillion-dollar wire fraud conspiracy and Venezuela sanctions evasion scheme (IRS Criminal Investigation) Source: Board chairman and co-owner of Puerto Rican bank sentenced in multimillion-dollar wire fraud conspiracy (IRS Criminal Investigation)​

The phone store worker who sold SIM swaps.

  • Kenneth Carter, a former AT&T store employee in Portland, Oregon, was sentenced on 8 September to 16 months in prison and ordered to pay $99,528 in restitution.
  • From May 2018 to November 2019, a co-conspirator sent him customers' personal details. Carter used his store access to move their phone numbers onto phones the group controlled.
  • That let the group receive the victims' bank reset codes and wire money out. Carter's cut was typically $1,000 to $2,000 per swap.
  • Three victims faced intended losses of at least $593,963. One lost $99,528 to a Portuguese bank account. Bank fraud controls stopped the other transfers.

Source: Oregon man sentenced to 16 months in federal prison for abusing his role at mobile phone store to give customers' info to criminals (IRS Criminal Investigation)​

The loan officer who advertised her access on Instagram.

  • Attallah Williams, of Hampton, Georgia, was sentenced on 1 September to six and a half years in prison and ordered to repay $3,272,000.
  • As a Small Business Administration loan officer, she approved fake pandemic grant applications from accomplices in return for a share. She recruited them on Instagram and promised referral fees for new recruits.
  • She then moved to the IRS as a tax examining technician and coached recruits through false Employee Retention Credit claims.
  • In total her scheme claimed more than $3.5 million.

Source: Former SBA and IRS employee sentenced to federal prison for taking bribes in $3.5 million COVID fraud scheme (IRS Criminal Investigation)​

All of these people had access with little oversight of what they did day to day. They were in trusted positions and used their insider access to exploit others and cause harm. We see this week on week. It's a good question to ask in your organisation. Where are these positions, and what harm could be done around finances?


AI Insider Threat Watch: The Agents Nobody Is Watching

  • Microsoft's 2026 Digital Defense Report says 88% of enterprises are already trying out AI agents. 82% of leaders plan wider rollouts in the next 12 to 18 months.
  • Agents pick up access over time that is rarely taken away. When an agent gains new skills, it may not trigger the access checks a staff member's change of role would. Many agents have no named human responsible for them.
  • Of the attack attempts Microsoft saw against AI agents from February to May 2026, 17% tried to pull out sensitive data. In its Azure AI telemetry, malicious link injection and exfiltration made up 52% of attack activity.
  • Microsoft's advice: "Each agent needs its own identity, not a credential borrowed from a human employee."

This is an issue we've been discussing for the last year, almost a year and a half, as agents have become more and more autonomous. Microsoft's suggestion is to give each working agent its own account. You still run into problems when agents use the user's privileges, or when their logging pathways are different. Then you can't see their chain of thought, the prompt handoffs, the responses or the agent-to-agent interaction. Getting visibility, just like in cybersecurity three decades ago, is still as important today, but with a focus on agent visibility.

Source: Insights from the 2026 Microsoft Digital Defense Report (Microsoft Security Blog)​


Mitigation Corner: The Laptops in the Spare Room

  • The US Justice Department announced coordinated action against North Korean remote IT worker schemes on 13 October. An Ohio indictment charges 14 people, 10 of them based in China. Two Americans were arrested: Michael James Bostelle of Pennsylvania and Juan Cecilio Pena Carrillo of Texas. They face wire fraud, money laundering and identity theft charges.
  • From 2019 to July 2025, North Korean IT workers allegedly used stolen or fake identities to get remote jobs. More than 70 US employers were hit. Helpers in the US hosted the company laptops at home and installed remote-access software. About $3 million is believed to have gone to sanctioned entities controlled by North Korea.
  • In September the FBI searched more than 40 Texas homes suspected of being laptop farms and seized about 187 laptops and KVM (keyboard, video, mouse) devices. The FBI says the workers are now recruiting more American helpers, and sometimes subcontracting whole jobs.
  • Mandiant's M-Trends 2026 found North Korean IT workers stayed undetected for a median of 122 days, and in several cases more than a year.

I think we're going to keep seeing evidence of DPRK activity for months, if not years, to come. We're still only finding the tip of the iceberg with the operators that are running the run-of-the-mill playbook. I think there are more advanced teams or individuals still in very key positions in organisations.

Check where new hires' laptops are shipped and where they log in from. A laptop sent to a different address, or remote-access tools installed in the first week, is worth a look.

Source: Indictment, United States v. Sun Sheng Ming et al. (N.D. Ohio) Source: Justice Department Announces Coordinated Actions to Combat North Korean Remote IT Worker Schemes Affecting More than 70 U.S. Companies (US Department of Justice) Source: FBI Public Service Announcement, 13 October 2026 (IC3) Source: M-Trends 2026 (Mandiant, Google Cloud)​


Trusted Profession Watch: The Banker and the Customer with Dementia

  • Carlos Bras, 41, a former Santander relationship manager at a branch in Seekonk, Massachusetts, was sentenced on 8 October to 30 months in prison. He pleaded guilty to mail fraud and aggravated identity theft.
  • His customer was a 78-year-old man with dementia who lived in assisted living and had a court-appointed conservator.
  • In 2023 Bras turned on online banking for the man's account, had cheques sent to addresses he controlled and ordered a debit card for himself. He moved money to accounts controlled by him and his family, and wired more to a bank account in Portugal. Investigators also found a Portuguese real estate contract signed in the man's name. His conservator said the signature wasn't his.
  • Santander's internal fraud team flagged him and called police, according to the Boston Globe. He was ordered to repay $126,000.

This is a great find by an internal fraud team, to catch this type of activity. And it's great to see that he's been ordered to repay the money he stole from someone who is vulnerable.

Source: Former Bank Employee Sentenced for Stealing More Than $125,000 from Elderly Customer with Dementia (US Attorney, D. Rhode Island) Source: Former R.I. bank employee accused of scamming elderly man reaches plea deal (Boston Globe)​


Trivia

This week's Headline is about a bug hidden in an executive meeting room. One of the most famous bugs in history was hidden inside a gift that Soviet schoolchildren gave the US ambassador in Moscow at the end of the Second World War. It hung in his residence for years before it was found in 1952. What was the gift?

A) A chess set B) A portrait of George Washington C) A carved wooden replica of the Great Seal of the United States D) A desk clock

(Answer at the bottom.)


Corporate News: Social Security's Insider Threat Program Falls Short

  • The Social Security Administration's inspector general found the agency's insider threat program "did not fully comply with Federal and Agency requirements." The report is dated 30 September. SSA holds personal information on nearly every US citizen.
  • Auditors found deficiencies in the program's policy, staff training, removing access for people who leave, documenting insider threat cases, and annual self-assessments.
  • Federal rules require a departing employee's access to be cut within a set time, and within 18 hours if their ID card can't be collected. A sample of departed staff showed deficiencies.
  • The details were withheld as too sensitive to publish. SSA agreed to all six recommendations.

I wouldn't be surprised if most organisations had gaps in their insider risk program, particularly government organisations. They're constantly underfunded, understaffed and generally playing catch-up with tech debt as well. We don't know how bad the gaps were in this example, but again, it's a continuous improvement program.

Source: Insider Threat Program, audit report 142302 (SSA Office of the Inspector General)​


Updates

Quick follow-ups on stories from earlier issues.

flydubai FZ1073. Following Issue #25 and Issue #26, the UAE's Attorney-General released new findings on 9 October. Prosecutors say the co-pilot planned to take control of the aircraft and crash it into the passenger terminal at Ben Gurion Airport. He allegedly used two earlier flights to Tel Aviv to learn the route and the airport. A pilot travelling on board noticed the cockpit door opening and closing, went in, and stopped the attack with help from crew and passengers. Prosecutors say he admitted acting alone, but they are still looking for possible accomplices. Source: Flydubai investigation: 7 findings that change what we know about attempted terror attack (Khaleej Times)​

ShinyHunters and the FBI. Following Issues #24 to #26, a third arrest. Edward Dubrovsky, a Canadian ransomware negotiator who co-founded the incident response firm CYPFER, was arrested in Pennsylvania on federal extortion and conspiracy charges. He has been transferred to the Eastern District of Texas. The complaint is sealed. The Boston Globe, citing people familiar with the case, reports that investigators see him as a primary co-conspirator in the attack on FBI data. The FBI has not named him. It's the second week running that someone paid to help ransomware victims has been accused of abusing that trust, after Issue #26's Zohar Pinhasi. Source: Cybersecurity executive who helped hacking victims arrested on federal extortion charges (Nextgov/FCW) Source: Cyber exec arrested in case allegedly tied to ShinyHunters hackers (BleepingComputer)​

Apple v OpenAI. We covered this in Issue #16, Issue #18 and Issue #21. On 13 October Judge Edward Davila ordered early discovery, citing a conversation between two former Apple employees about wiping their Apple-owned devices. At a hearing on 14 October he said he was "highly likely" to grant Apple a preliminary injunction. OpenAI then agreed not to use the disputed information and to collect forensic images of 14 employees' devices, a deal the court can enforce. OpenAI says it has "no need or desire for Apple's trade secrets." Source: Apple, OpenAI head to court in high-stakes AI trade secret fight (Bloomberg Law) Source: OpenAI agrees not to use alleged Apple trade secrets amid ongoing lawsuit (9to5Mac)​


What's On

iTA Workshop: The Insider Attribution Paradox, a DTEX Insider Threat Advisory workshop. A reminder from last week, and it's next week.

  • When: Tuesday 27 October 2026, 1:00 PM to 2:00 PM PT (Wednesday 28 October, 6:00 AM AEST Brisbane time).
  • What it covers: how AI agents embedded in enterprise workflows complicate attribution, what OpenTelemetry can reveal about agent workflows and tool execution, closing the gap between user, agent and tool, and investigating AI-assisted activity.
  • Online, and open to the wider insider risk community.
  • Register: https://www.dtex.ai/events/ita-workshop-the-insider-attribution-paradox/​

Insider Risk Community of Practice Networking Drinks, Melbourne, hosted by the Australian Cyber Collaboration Centre. Wednesday 11 November 2026, 5:15 PM to 7:00 PM AEDT, SEEK Head Office, Cremorne. Register: https://events.humanitix.com/insider-risk-community-of-practice-networking-drinks-melbourne-november-2026​

On the horizon: the Critical Pathway to Insider Risk Conference at Fordham University, New York, on 4 and 5 November (https://www.insiderriskgroup.com/), and the Australian Insider Risk Symposium in Sydney, a half-day event in November with the date still to be confirmed (https://www.irpa.org/event-details/australian-insider-risk-symposium).


Resources

A few places worth your time if you want to go deeper on insider risk.

Worth bookmarking:

And my own: Insider Threat Awareness Training on YouTube, a free course with certification.


Trivia Answer

C. A carved wooden replica of the Great Seal of the United States. The bug inside, nicknamed "The Thing", had no battery or wiring. It only worked when the Soviets beamed a radio signal at it from outside, which made it very hard to find. It was discovered in 1952 in the ambassador's study at Spaso House, his Moscow residence. A replica is on display at the NSA's National Cryptologic Museum.

Source: Cold War: The Great Seal (National Cryptologic Museum, NSA)​

Insider Threat Weekly

*make sure to check your spam folder for the signup acceptance email.