Anthropic Grew Counterintelligence Roles 2.4x to 33 in 2026

Anthropic counted 33 counterintelligence-type roles so far in 2026, up from 14 in 2025 and nine in 2024, according to Revelio Labs data published by Reuters on 12 November and reported by Fortune on 14 November. That is 2.4 times last year’s count and 3.7 times the 2024 figure, by our arithmetic.

Reuters’ chart counts job adverts related to security-clearance and insider-risk jobs; Fortune describes the same figures as hires. OpenAI’s count rose to 22 from nine, also 2.4 times, and Anthropic’s 33 equals about 0.6% of the approximately 5,300 employees its S-1/A reported at 30 September, by our arithmetic.

Business Pill · Least Privilege

A short explainer on least privilege: giving a system or person only the access its task needs. It teaches the general idea and does not discuss Anthropic, its hiring or the firms in this piece.

What the Revelio Count Shows

Reuters reported that “AI firms have been ramping up hiring of roles that have a broad counterintelligence function”. Revelio Labs told Reuters that counterintelligence hiring was a new phenomenon at frontier AI firms, and that most salaries ranged from $200,000 up to around $500,000.

In 2026 Anthropic’s 33 is 1.5 times OpenAI’s 22, by our arithmetic. Reuters reported that Anthropic added former CIA deputy director David Cohen and former NSA cybersecurity director Dave Luber to the national security advisory council it launched in August 2025, and that OpenAI appointed former NSA director Paul Nakasone to its board in June 2024.

An Anthropic advert for a threat-intelligence security engineer describes the company as “one of the most interesting targets in the world for nation-state and advanced criminal actors”, Reuters reported. On 10 November Anthropic posted a Security Engineer, RF and Technical Surveillance Countermeasures role at $320,000 to $405,000 a year, covering “on-site RF monitoring and detection, technical surveillance countermeasures (TSCM) and emanations security” at its data center sites, the posting says.

What Was Posted on X

The account @choblin29 posted this on 15 November, summarising two outlets: Interfor’s CEO telling the Guardian about around a dozen investigations in three years, and, citing Fortune, that Anthropic hired 33 counterintelligence staff this year, up from 14 in 2025. The piece traces the 33 to Revelio Labs data published by Reuters, notes that Reuters’ chart counts job ads while Fortune calls them hires, and sets the count beside Anthropic’s own S-1/A.

What Anthropic’s S-1/A Says It Protects

The S-1/A Anthropic filed on 10 November defines model weights as “the primary intellectual asset of a frontier intelligence developer, representing the accumulated investment in compute, data, and research”. Our Anthropic IPO map tracks the offering itself.

The filing says those weights and other core intellectual property “represent particularly high-value targets for state-sponsored espionage and theft”, and that an unauthorized exfiltration “could enable adversaries to bypass significant research investment, eliminate key competitive advantages, and potentially raise national security concerns”.

On insider risk it says: “We have faced and expect to face insider threat risks and vulnerabilities related to hiring processes, identity verification, remote work environments, and compromised devices that could enable unauthorized access to sensitive systems or information”.

It adds that staff with authorized access may be compromised after joining, including through efforts “to recruit, bribe, coerce, or otherwise induce such personnel”. And it says “addressing these risks also requires us to attract and retain additional specialized security personnel”.

The Cost Shows Up in Pay and in the Schedule

Kathy Lavinder, founder of recruiter SI Placement, told Reuters that the pay, which Reuters described as well above the public sector, reflects the relative scarcity of talent, and that top corporate security chiefs can earn $1 million to $4 million a year, depending on stock awards. We covered OpenAI’s security investigator posting on 12 November.

Christopher Falkenberg, president of Insite Risk Management and a former US Secret Service agent, told Reuters: “Demand outstrips the supply of people with relevant experience”.

The S-1/A also puts a schedule cost on security: “we recently chose to delay pre-training a new frontier base model for more than a month to allow time to further enhance our security measures related to preventing model weight exfiltration and protecting our intellectual property”.

It says Anthropic has “operated in the past, and may in the future operate, for periods of time with known limitations or gaps in our defenses as we continue to mature our security posture”.

What Security Firms Say Companies Face

Don Aviv, CEO of the corporate intelligence firm Interfor International, estimated that his firm has conducted “around a dozen investigations for significant companies” over the past three years into schemes involving honeypots, coercion or blackmail, the Guardian reported on 15 November.

Catherine Marinis-Yaqub, a former CIA officer at Control Risks, told Reuters that companies value counterintelligence experts for spotting threats such as employees stealing trade secrets and attempts to exploit vulnerabilities, including via executives, by blackmail or coercion.

Eduardo Jany, head of global security for a major US corporation, told the Guardian: “It’s really more about social engineering”. Outside AI labs, Reuters reported that ASML has set up an Intelligence Fusion Centre and is advertising for an intelligence officer based in Taiwan with at least 10 years of counterintelligence experience.

The Structural Read

As we read the sources, the cost of protecting weights shows up in two places: pay for scarce specialists, with Anthropic posting $320,000 to $405,000 for a surveillance-countermeasures engineer, and time, with the S-1/A describing a frontier pre-training run delayed by more than a month for security work.

As we read the Revelio counts, OpenAI’s rise to 22 from nine is the same 2.4 times as Anthropic’s, by our arithmetic, so the data show the same multiple at both labs.

Three Implications

FOR THE IPO Anthropic lists model-weight exfiltration and insider threat among the risk factors in its S-1/A, filed on 10 November.

FOR THE TALENT MARKET Revelio Labs puts most counterintelligence salaries at $200,000 up to around $500,000, Reuters reported; Kathy Lavinder told Reuters top security chiefs can earn $1 million to $4 million a year.

FOR THE RATIO Anthropic’s 33 roles equal about 0.6% of its approximately 5,300 employees, by our arithmetic; the two figures differ in kind and date.

The Business Engineer Lens

This story maps onto the Business Engineer framework Inside Anthropic’s Permission Layer.

The framework says of the governance layer: “It is where decisions are made about who gets access to frontier intelligence, under what conditions, and for which purposes.”

As we read it, counterintelligence and insider-risk staff enforce that access inside the lab itself, around the model weights. The framework predates these counts and does not discuss them.

What Is Not Established

Reuters’ chart note describes the Revelio figures as the number of ads per year related to security-clearance and insider-risk jobs, while Fortune describes them as hires. Which roles count as counterintelligence is Revelio’s classification, and the 2026 figure covers only part of the year.

The 0.6% ratio sets a 2026 count against a 30 September headcount, so it is not a measure of the size of Anthropic’s security team. Interfor’s figure is its CEO’s estimate for his own firm. We did not contact Anthropic for this piece.

The Bottom Line

Anthropic’s counterintelligence-type roles rose to 33 in 2026 from 14 in 2025, 2.4 times, by our arithmetic, per Revelio Labs data published by Reuters. Its S-1/A calls model weights the primary intellectual asset of a frontier developer and says it has faced insider threat risks.