AI-generated illustration

A US cybersecurity startup only a few months old, Cathedral, reportedly found its way into Anthropic's sensitive internal data and messaging systems in less than 12 hours, according to a draft report the company prepared.

Anthropic confirmed that a limited portion of its sales, human resources and IT systems was accessed and said federal officials were informed. Separately, the Claude developer filed paperwork with US regulators for an initial public offering.

How the breach was carried out

The Washington Post obtained a draft report prepared by Cathedral, a cybersecurity firm founded by former employees of the Department of Government Efficiency (DOGE), the unit Elon Musk previously worked in. The company sells AI-powered cybersecurity tools to the US government.

According to the report, AI agents inside Cathedral found a route into Anthropic's internal communications app in under 12 hours. The app contained meeting transcripts, salary information, job postings and messages about the company's advanced artificial intelligence models.

Anthropic confirms limited access

Anthropic said in a statement that Cathedral accessed a "limited portion" of its sales, human resources and IT systems, and confirmed that federal officials had been briefed on the incident.

Washington official raises security concerns

A US administration official, speaking on condition of anonymity, said the report had circulated within the government and raised "national security risk concerns." The official said that "if three or four kids can start a company and do this, the Chinese are probably not far behind."

Anthropic files for a US IPO

Anthropic, the developer of the Claude AI models, filed a registration statement with the US Securities and Exchange Commission (SEC) for an initial public offering. The filing said the company plans to list its Class A common shares on Nasdaq under the ticker "ANT."

The number of shares to be sold and the expected price per share have not yet been disclosed. Anthropic and existing shareholders both plan to sell stock in the offering, though the company will not receive proceeds from the shareholders' sales. The listing is planned to start as soon as possible after the registration statement takes effect.

The filing names 22 financial institutions involved in the offering, including Goldman Sachs, Morgan Stanley, J.P. Morgan, Citigroup, Barclays, Wells Fargo Securities, Bank of America Securities and Deutsche Bank Securities.

Revenue and losses disclosed

The filing also includes financial results. Anthropic reported revenue of $16.3 billion in the first six months of 2026, alongside a net loss of $80.7 billion. The company's net loss was $41.9 billion in 2025, $8.3 billion in 2024 and $1.5 billion in 2023.

The US-based AI company said in June that it had submitted a confidential draft registration statement to the SEC as part of the planned offering. International media reports indicate the company is aiming to carry out the IPO in November.

Key takeaways

• Cathedral, a months-old firm started by ex-DOGE staff, accessed Anthropic data in under 12 hours
• Meeting transcripts, salary data and messages about advanced AI models were exposed
• Anthropic confirmed limited access to sales, HR and IT systems and notified federal officials
• Anthropic filed for a Nasdaq IPO under the ticker "ANT", reporting $16.3 billion in revenue in the first six months of 2026