Anthropic spent $2.75 million protecting cofounder and CEO Dario Amodei in 2025—more than four times his $654,167 base salary—according to the AI company's newly public IPO filing.

The security bill accounted for more than 15% of Amodei's $17.96 million in total reported compensation for 2025, a figure that also included stock and option awards.

Amodei was not the only executive Anthropic paid to protect. The company spent another $1.29 million on personal protection, home security, and transportation for his sister—president and cofounder Daniela Amodei—plus $97,055 on private-aircraft travel for Dario that the company said was required under his security program.

Anthropic has maintained round-the-clock security since 2024, the Wall Street Journal reported in July. According to the S-1, Anthropic's board determined in 2025 that enhanced protection for the Amodeis and their families was necessary, citing security studies, increased media attention, and what it called "bona fide business-related, heightened security concerns." The measures included personal protection, home surveillance, and private cars with drivers.

Anthropic did not disclose how much it spent on executive security in 2024, making it impossible to determine whether costs rose sharply in 2025. The filing acknowledges threats to employee safety but does not specify when they occurred.

"Heightened public attention on AI has increased physical and informational security risks," the company wrote in the filing, "and we have experienced threats to employee safety and attempts to obtain confidential information through physical or digital means."

Threats against Anthropic have spilled into public view

Several of those threats became public this year. In April, days after an attack on OpenAI CEO Sam Altman's home, a man slipped into the lobby of Anthropic's San Francisco office behind a badge-swiping employee and told a security guard that a top executive was "going to be killed," according to the Journal.

That same month, a man who had applied for a job at Anthropic under a false identity allegedly posted online threats to harm employees' children, framing them as punishment for the company stealing his work. Police classified the post as a terroristic threat but made no arrest. In June, a customer demanding a refund wrote that he would come to Anthropic's office with his pistol.

In August, Amodei himself was named in an alleged threat. A Claude user claimed in a chat to have purchased an AR-15 rifle and to have the CEO "in his sights," according to a police report obtained by the San Francisco Standard. Anthropic reported the chat to police and banned the account. The man later told the Standard he had been joking and was embarrassed.

The growing threats facing AI executives

Digital threats targeting AI executives and data centers surged sevenfold from late February to May 2026, according to security intelligence firm Liferaft. "What has surprised me is how bad it's gotten over such a short period of time," Liferaft CEO Jonathan Graff told the Journal.

The threats have also moved offline. In April, a Texas man was charged after allegedly throwing a Molotov cocktail at Altman's San Francisco home, then going to OpenAI's headquarters, where he threatened to burn down the building. Federal prosecutors said he was carrying an anti-AI document that advocated killing AI executives and investors and listed names and addresses. He has pleaded not guilty.

How Amodei's security bill stacks up against Zuckerberg's

Anthropic said in its IPO filing that the security measures it provides senior executives are comparable to those at other high-profile technology companies. By that measure, Amodei's protection is far from the priciest in tech.

Meta remains in a league of its own. The company disclosed approximately $8.5 million in direct security costs for CEO Mark Zuckerberg in 2025, plus a $14 million annual allowance for additional protection for him and his family—about $22.5 million in total, or $25.1 million including private-aircraft costs. That is more than eight times what Anthropic spent on Amodei. Last year, Fortune reported that Meta spent more guarding Zuckerberg than Apple, Nvidia, Microsoft, Amazon, and Alphabet did for their own CEOs—combined.

Here is how disclosed CEO security spending compares:

• Meta / Mark Zuckerberg: About $22.5 million, including a $14 million allowance ($25.1 million including private-aircraft costs)
• Alphabet / Sundar Pichai: $8.8 million
• Tesla / Elon Musk: About $4.8 million paid to a security company owned by Musk, which Tesla says covers only a portion of his total security costs
• Nvidia / Jensen Huang: Nearly $4 million in fiscal 2026, which ended Jan. 25, 2026
• Anthropic / Dario Amodei: $2.75 million, plus $97,055 for security-related private-aircraft travel
• Amazon / Andy Jassy: About $1.7 million for security and business travel combined (Amazon separately paid $1.6 million for Jeff Bezos's security and business travel)
• OpenAI / Sam Altman: Not publicly disclosed

The blank entry for OpenAI underscores how these figures usually come to light: SEC rules require registered public companies to report executive pay, including security-related benefits, while private firms face no comparable obligation. Itemized numbers like Anthropic's are a byproduct of the disclosure process the company is now entering through its IPO filing.

Compared with the multimillion-dollar security bills of other high-profile tech executives, Anthropic's spending may not stand out. Against the broader corporate world, however, the gap is stark.

ISS-Corporate, a corporate governance research firm, told Fortune that a recent study shows the median reported CEO security cost was $93,818 among S&P 500 companies and $401,336 among software and services companies. Amodei's $2.75 million security bill was nearly seven times the latter figure.

Valeriano Saucedo, an associate director at ISS-Corporate, said the better comparison may have less to do with the company itself and more to do with the person running it: How high-profile is the CEO?

When corporate backlash becomes personal

AI promises to transform the world, but even some of the people building it warn that it could destroy humanity. In September, Anthropic alignment science lead Evan Hubinger said he personally believed there was a greater than 10% chance AI could kill all humans within the next decade. Anthropic's own IPO filing warns that advanced AI systems could pose "catastrophic or existential risks to humanity" without adequate safeguards.

As AI companies become the focus of heated public debate, their CEOs are also becoming the faces of the technology.

Meta has long warned that public anger toward a company can become personal. In its proxy filings, the Facebook parent has described Zuckerberg as "synonymous with Meta," saying negative sentiment toward the company is often transferred to him. The 2024 killing of UnitedHealthcare CEO Brian Thompson showed how corporate leaders can become targets amid broader resentment toward their industries, and prompted many companies to reexamine how they protect their top executives.

Executive security benefits are becoming more common. According to ISS-Corporate, 28.3% of S&P 500 companies reported providing security benefits to their CEOs in 2025, more than double the 13.4% in 2021. Technology has seen a particularly sharp rise: A separate study of the 500 largest U.S. public companies by executive data firm Equilar found the tech sector posted the biggest increase in the prevalence of executive security benefits, up 73.5% between 2020 and 2024.

Who decides how much security is enough?

Anthropic's security program has since expanded beyond its founding siblings. In 2026, the company began providing security services to CFO Krishna Rao and commissioned an independent assessment of the risks he faces. It has not disclosed how much that protection costs.

ISS-Corporate said boards should consider whether security expenses address legitimate business risks, whether the costs are reasonable, and whether the arrangements are regularly reassessed. Anthropic said in its filing that its board regularly reviews the security measures provided to the Amodeis and their costs.

The firm added that private-aircraft use can be justified as part of an executive security program, but boards should also consider whether less costly alternatives would be appropriate.

"The dollar amount alone does not determine whether an expense is appropriate," Saucedo said.

If Anthropic completes its planned listing, annual SEC filings would build a public year-over-year record of these costs, giving shareholders and proxy advisors like ISS-Corporate the kind of regularly reassessed documentation its governance guidance calls for.

This story was originally featured on Fortune.com.