These are the AI companies that the CIA is investing in
You can’t spell CIA without AI
A brave anonymous source has leaked a top-secret list of AI companies that the CIA has been pouring millions of investment dollars into. Just kidding… it’s all publicly available on their website.
In-Q-Tel is the non-secret, non-profit investment arm of the CIA, which was established in 1999. A 501(c)3, the firm is not officially part of the US government (though it is largely controlled by it), and has received over $1.2 billion from US taxpayers since 2011, which it has used to make over 750 investments. In-Q-Tel’s mission is to "Be the premier partner trusted to identify, evaluate, and leverage emerging commercial technologies for the US national security community and America’s allies.”
Rather than seeking financial reward for backing winning companies, In-Q-Tel’s investments are strategic — it surveys the marketplace for technologies and businesses that could solve problems for the US intelligence and national security community.
A most unusual venture capital firm
While it does not disclose all of its investments, a look through its publicly disclosed portfolio shows that In-Q-Tel made some smart, early bets. Some high-profile companies In-Q-Tel invested in include Keyhole, the satellite mapping app which Google later acquired and turned into Earth, and Palantir Technologies, the data analytics firm co-founded by Peter Thiel, currently valued at roughly $80 billion. The firm has since exited those positions.
In-Q-Tel is a huge player among venture capital firms investing in national security companies. The Silicon Valley Defense Group’s NATSEC100 index is a ranked list of the top-performing, venture-backed private companies working in the national security sector. This year’s list ranks In-Q-Tel as the number one venture capital firm in the space, having backed 35 companies on this year’s NATSEC100 list.
In a 2020 interview, George Hoyem, Executive Vice President for investments at In-Q-Tel said the company makes 50-60 investments per year.
“95% of the companies we back have never done business with the federal government, so we’re taking commercial startups and showing them a dual use into the government,” said Hoyem.
Hoyem said the typical investments are in the $1-4 million dollar range, with smaller equity-only seed investments of around $1 million or less, while “work program” deals where the company’s technology can be developed and tested by partner defense and intelligence agencies see a larger investment. According to Hoyem, 70% of their investments make it to the pilot stage, and about half are adopted for actual agency use.
Hoyem said that they look for new technologies that their national security and intelligence partners may not be aware of. “So we’ll go back to them and say, ‘Hey I know you’re looking at this problem this way, but there’s this new capability’ whether it be, you know, an accelerated GPU engine, or artificial intelligence being applied to a particular problem,” Hoyem said.
Some of the companies that In-Q-Tel has invested in build technology that could have come straight out of the labs of “Q,” James Bond’s spy gadget wizard, which was an inspiration for the company’s name. The company has backed mouth radios, a Brooklyn hologram studio, and skin-care products that could be used to extract your DNA.
Unlike the CIA’s financials, which are classified, In-Q-Tel is a non-profit, which means we get a peek into how it invests its funds through its public tax returns. Looking back over the past decade or so, In-Q-Tel’s net assets grew substantially year-over-year.
In 2011, In-Q-Tel reported around $100 million in net assets, growing to $930 million by the end of 2023. Even though it is a non-profit, In-Q-Tel still pulls in substantial revenue from investment returns in its portfolio, which it is free to use to re-invest in more companies.
In the past five years for which tax returns are available, the US government has granted In-Q-Tel around $100 million per year. While the company’s assets have grown, it is hardly self-sustaining and requires substantial cash infusions from US taxpayers to stay afloat. If you remove government grants from In-Q-Tel’s revenue numbers, after expenses they would have had a loss for the past 12 out of 13 years.
Questions about transparency
The clandestine nature of the CIA may have rubbed off on the culture of In-Q-Tel. A 2016 investigation by The Wall Street Journal highlighted the lack of transparency at In-Q-Tel, and identified connections between In-Q-Tel’s trustees and the boards of the companies that received investment funds, prompting concerns of conflicts of interest arising from the use of taxpayer funds.
In-Q-Tel – also known as the more Silicon Valley friendly moniker IQT– lists a sampling of the companies it has invested in across space, life sciences, energy, quantum computing, and of course AI. In fact, AI makes up the largest share of In-Q-Tel’s active investment portfolio.
You can’t spell CIA without AI
Drilling down into the AI slice of the portfolio gives us a fascinating glimpse into the future strategic bets of one of the most secretive parts of the US intelligence community. We looked through the AI companies in this portfolio, to see what types of applications the intelligence agency is nurturing.
The largest slice of investments were made in 16 companies that host AI infrastructure for hosting or deploying models, making up about 28% of the AI portfolio. In 2016, In-Q-Tel made one of its most successful investments in Databricks, a data warehousing and AI platform that is currently valued at $43 billion, which pulled in $1.6 billion in 2024 alone.
Dual use is key
Most of the companies in In-Q-Tel’s portfolio offer “dual use” technologies that might help a business gain a competitive edge in the corporate world, but could also be used by a nation state to dominate its enemies.
For example, technology that helps with understanding languages and the nuances of conversation is obviously something that the intelligence community would be interested in (think monitoring the conversations of our enemies at scale). This is also true for customer service companies, who might want to monitor the emotional stress levels in an angry customer’s voice to escalate a call to a human operator. This is just the kind of dual use technology that Hoyem said the company is looking for in the marketplace.
In 2023, In-Q-Tel invested in Behavioral Signals, a Los Angeles based company that sells a variety of speech and conversation based tools for the customer service, sales, and defense industry. On their website, the company offers its platform to “utilize vocal analysis to decipher emotions, intentions, and stress levels in any communication. With unparalleled precision, gather intelligence, identify threats, and screen personnel.”
On a page detailing defense and intelligence applications for the company’s platform, it advertises “Polygraph Plus,” which can use AI to “assess a candidate or threat before polygraph testing.” The page also describes how the company’s tools can be used to “Predict intent (do harm or no harm), detect stress or duress, detect fraud, detect ’control’ and trustworthiness.”
Another dual use company In-Q-Tel’s AI portfolio is Fiddler.AI, a predictive AI platform that was founded “to bring trust and transparency to AI everywhere,” according to the company’s website. A page detailing the company’s commitment to “responsible AI” says “We provide comprehensive tools that shed light on bias to develop responsible AI that is fair and inclusive in the hope of building a better world with less discrimination.”
But elsewhere on their site, the company advertises a less noble vision. Its predictive AI models which power “autonomous vehicles, such as aerial drones and unmanned underwater vehicles” to analyze imagery “to anticipate potential threats and navigation.” The company also promotes the ability to monitor sensor data to “Improve target precision with high performance model monitoring across different battlefield environments.”
Another large group of investments includes eight companies that sell geospatial and remote sensing tools, such as satellite imagery analysis and monitoring.
Intelligence agencies used to regularly monitor troop movements and submarine construction sites with film-based satellite imagery, but now companies can also get in on the game, counting cars in parking lots, solar panels on homes, or the quality of crops. Pairing this extremely useful imagery with AI, companies can extract meaningful, valuable data from the surface of the earth, another great dual-use example.
In-Q-Tel made such an investment with Blackshark.AI, based in Graz, Austria. The company advertises that its “synthetic 3D globe” technology was used in Microsoft Flight Simulator to create the game’s impressive photorealistic landscapes.
But when gamers grow up and start working for an intelligence agency like the CIA, they can also use Blackshark.AI’s tech to find the bad guys using the company’s “ORCA HUNTR,” which gives customers “the unique power to identify any object on the Earth’s surface with unprecedented ease and precision. “What if you could detect any object on the surface of the planet? What would you detect?,” reads their website.
Culture clash
While In-Q-Tel may have proven that it can successfully identify and back promising start-ups with its deep pockets full of taxpayer dollars, the geographic and cultural divide between the DC beltway and Silicon Valley is great, and it may be keeping the two worlds from a deeper relationship.
In testimony before a 2020 House Intelligence subcommittee hearing discussing technological innovation in the intelligence community, former In-Q-Tel CEO Chris Darby suggested that more work was needed to bridge this divide.
Darby said that when a four-star general representing the Department of Defense visits companies in Silicon Valley, “there are 50 people trailing behind. That is not the way the valley operates. The valley operates in a very casual mode and it is not about giving a written speech in front of an audience, it is about having a conversation in a cafeteria.”
“Startups don’t speak government, and government doesn’t speak start-up,” explained Darby.
But one thing they have in common: They love pouring money into AI.
In-Q-Tel declined to comment for this story.