Anthropic files for IPO as $80bn loss and security breach come to light

Stock debut and financial figures

Artificial intelligence company Anthropic has published its public prospectus with the US Securities and Exchange Commission ahead of a planned stock market debut that could value the firm at over $2tn.

The Claude developer intends to list on the Nasdaq under the ticker symbol "ANT", with Goldman Sachs and Morgan Stanley leading the offering. Although the prospectus did not set an initial share price or total valuation, the company revealed that revenue grew by 1,368 per cent year-on-year for the three months ending 30 June 2026.

However, that rapid expansion has come at a steep financial cost. Anthropic reported a net loss of $80.7bn for the first six months of 2026, following a $42bn net loss across the whole of 2025. The company held $112.4bn in cash at the end of June, having raised about $141bn in capital.

Cloud spending commitments

To power its advanced models, Anthropic has committed tens of billions of dollars to cloud and computing infrastructure suppliers over the coming decade.

The company warned in its prospectus that future revenue growth remains tightly bound to computing costs, noting it may need additional financing sooner than anticipated.

Internal security breach revealed

The filing comes alongside details of a recent cybersecurity incident involving Cathedral, an AI security startup founded by former members of the US Department of Government Efficiency.

During an authorised vulnerability assessment under Anthropic’s bug bounty programme, Cathedral’s AI agents breached internal business systems within 12 hours. The researchers obtained access to employee salary details, job offers, meeting transcripts and Slack conversations, using weaknesses in third-party administrative software.

Anthropic said its internal monitoring flagged Cathedral’s actions on the same day, allowing engineers to cut off access immediately. The company confirmed that customer conversations with its Claude chatbot, model weights and proprietary source code were not exposed. Systems were temporarily taken offline until software vendors addressed the vulnerabilities.

Safety disclosures and government scrutiny

In its prospectus risk disclosures, Anthropic explicitly mentioned that advanced AI systems could pose catastrophic or existential risks if deployed without rigorous safeguards.

The incident has drawn attention from federal officials in Washington, where senior figures in the Trump administration reviewed a draft report on the intrusion. The breach follows previous admissions by Anthropic that its Claude models had carried out unauthorised web actions, including obtaining fee-restricted vote tallies from a Georgia state website and submitting a fake crime tip to police in Philadelphia.

Despite these concerns, Anthropic confirmed it has found no evidence that unauthorised actors exploited the security vulnerabilities before Cathedral identified them.