Head of CIA OSINT emphasizes centrality of Twitter and Telegram surveillance

During a live webinar this afternoon, CIA Open Source Enterprise director Randy Nixon dodged questions on the CIA's use of commercial data and explained usage of Twitter for "tactical" surveillance.

A screenshot from a live interview with Randy Nixon, director of the CIA’s Open Source Enterprise, hosted by The Cipher Brief on September 28, 2023. Nixon conducted his interview from his parked truck and joked that he was sweating because of the heat, not purely because of the questions. The author’s first question, on the ethics of U.S. intelligence agencies purchasing commercial facial recognition from firms such as Clearview AI, is fully visible on screen.

The director of the Central Intelligence Agency’s unit for scouring the internet and hoovering up databroker profiles, Randy Nixon, is frustrated by both the “simplicity” of Bloomberg’s recent coverage of his agency’s capabilities and by Hollywood’s focus on spies. During a panel hosted this afternoon by the spycraft social network and newsletter The Cipher Brief, Nixon lamented that “People forget about [Open Source Intelligence] more often than not, because HUMINT [Human Intelligence] gets all the Hollywood attention. You know, you can watch any TV show and it’s all about HUMINT.”

Nixon was publicly revealed as the head of the CIA’s Open Source Enterprise (OSE) — which lives within the agency’s Directorate of Digital Innovation — through a LinkedIn post in December. Nixon conducted the interview from the back seat of his parked truck and was quick to joke that “If people start to see me sweat, it’s my location, because I bounce between buildings all day long and I’m actually sitting in my truck. It’s not the questions, or maybe it’s not only the questions.”

Nixon primarily criticized Bloomberg’s coverage of OSE’s new tools — and “the thousands of places that picked that up” — for having been “a little bit simplistic and [focusing] very much just on one feature of our technology, the chat, which is in beta. But what it misses is that [Artificial Intelligence], machine translation, AI-enabled other tools, including summarization, sentiment, what not, has been a complete change to the OSINT [Open Source Intelligence] community.” Throughout his interview, Mr. Nixon made repeated reference to the previous baseline being Next Generation Trident, a CIA analytical tool developed more than a decade ago.

As Nixon tells it, artificial intelligence text summarization has cured his unit’s “greatest weakness”, that it collects more surveillance data than any human has time to manually analyze:

“When we started in the ‘40s it was just newspapers, and radio in a couple of countries we cared about. The ‘50s came along and blew our minds, now we had television we had to worry about along with the Soviet bloc which had us covering the world. And you just keep marching through time and our job just keeps getting harder and harder. TV to cable TV to news…and the nascent internet and…social media and big data. You just keep going through and my officers keep getting overwhelmed…And we are delivering millions of pieces of information every day. Thousands and thousands of hours of video…there are times where there are analysts that will filter out the OSINTs that we’re delivering into their search engine and then only look at the HUMINT, SIGINT [Signals Intelligence], and what not.”

A similar narrative fueled the Pentagon’s controversial “Project Maven” effort for processing the high-resolution “Gorgon Stare” surveillance footage from MQ-9 Reaper drones with computer vision built in partnership with companies such as Google, Microsoft, Amazon, Clarifai, and Palantir. As explained by the former director of Project Maven, Jack Shanahan, as part of an interview with defense venture capital firm Lux last month: “This particular sensor that went on one of these unmanned aerial vehicles, commonly known as a drone, basically this thing could park over an entire city and look at the entire city 24/7. No analyst on the face of the planet could be able to analyze that information…which is why we need Google’s software engineers.”

While Palantir is largely understood to have taken over Project Maven, the U.S. tech giants Google, Amazon, Microsoft, Oracle, and IBM were the winners of the CIA’s mutli-billion dollar “Commercial Cloud Enterprise” — or C2E — cloud contract. And Nixon namechecked Google’s machine translation capabilities as the gold standard within the CIA’s open source work as part of his interview earlier today.

Despite the marketing appeal of “open source” intelligence, the methodology includes the usage of numerous controversial commercial surveillance technologies, ranging from scraping billions of photos from social media sites to build facial recognition software such as Clearview AI and PimEyes, to secretly harvesting the GPS data of potentially billions of phones through backroom deals with app developers and advertising data brokers, especially by way of companies such as Babel Street, Anomaly Six, and Outlogic. Nixon cheerfully noted that, since the CIA began incorporating automatic summarization technology, “The only thing that holds [us] back on collection is really … having the amount of money to go out and buy everything that’s out there.” (Mr. Nixon further stated that, rather than just working with the traditional “four or five” major defense contractors, more than 300 companies have become involved in the CIA’s “Digital Hammer” Broad Area Announcement, which he advertised on his LinkedIn roughly one year ago. Nixon also repeatedly endorsed Janes, the sponsor of his interview, for the company’s utility in studying weapons systems.)

A screenshot of the author’s interface to today’s live interview with Randy Nixon, the director of the CIA’s Open Source Enterprise. The author’s second question, on the ethics of U.S. intelligence agencies purchasing commercial cellphone location-tracking data, is fully visible on screen.

The manner in which the author’s questions on the Open Source Enterprise’s potential purchase of controversial surveillance technologies were diverted by both the moderators and Nixon is instructive. Despite my inquiries asking about usage of particular facial recognition products — such as Clearview AI, PimEyes, and Paravision — as well as about particular cellphone location-tracking databrokers — such as Babel Street and Outlogic — the questions were combined and watered down by moderator Suzanne Kelly to “potential ethical considerations of using Publicly Available Information, such as some international private sector cellphone type data”.

Nixon responded by lamenting that “Congress is very involved and they ask this very same question a lot”, but he noted that he couldn’t “go into minute details, because then it stops being unclassified.” The OSE director further noted that “If any spidey sense ever goes off on some weird thing we’re trying to do to get collection, the first thing I do is bring in my lawyer and the counter-intelligence officer so we all walk through it all together.” Concerning abuses of commercial surveillance data by U.S. intelligence agencies, including the surveillance of location data through a Muslim prayer app by the company now known as Outlogic, were acknowledged in a widely discussed declassified report to the Office of the Director of National Intelligence written in January 2022. But the author documented that Clearview AI, Outlogic, Babel Street, and Anomaly Six all sent in-person attendees to Naval Special Warfare Command’s Trident Spectre Industry Day earlier this month. And Nixon’s response focused almost exclusively on legality rather than addressing broader ethical concerns.

Despite pointedly avoiding discussion of the CIA’s potential relationship with controversial databrokers such as Clearview AI and Babel Street, Nixon was remarkably candid about how central the CIA’s surveillance of social media platforms Twitter and Telegram is to the agency’s open source efforts:

“I’ll give you a fictional example if that’s okay with you, because if I gave a real example, that source might disappear tomorrow or move from publicly available to commercially available, and we only have so much money. So, in a fictional example that’ll show you how we use it, so let’s imagine that there is a warlord threatening a coup in some fictional far-off nation. There’s our scene. So I have not said any particular country or region…OSINT is almost always the first tip I have that there’s a problem…today, that first hint is often somebody tweeting, or on Telegram…Let’s further imagine that then there’s a tweet or something that alerts us to who…this warlord is, or the people around him, we’re going to dig in very tactically about who the person is, where they like to be, you name it, what kind of juice they like. Anything you can get. And that kind of information is going to be very valuable to the warfighters and the operators who would also be working on this topic.”

But Randy Nixon and the CIA’s Open Source Enterprise seem to welcome even controversial press coverage. As part of a critique of proposals to use the investigatory nonprofit Bellingcat as a model for a new “open source” U.S. intelligence agency, Nixon quipped that “I love that people talk about [Open Source Intelligence], because it keeps us in the press. And…the public, Congress, others, understanding OSINT better is very helpful to me.”