Cybersecurity startup Cathedral claims AI agents breached Anthropic systems as AI maker files for IPO
Months-old startup Cathedral says its AI agents penetrated Anthropic's business systems in under 12 hours through third-party software, coinciding with Anthropic disclosing model training security pauses in its IPO filing.
Cathedral, a cybersecurity startup founded by former DOGE staffers, reported that its researchers used autonomous AI agents to penetrate Anthropic's internal business systems in under 12 hours. According to a draft report reviewed by The Washington Post, the agents exploited vulnerabilities in third-party software to access Anthropic's sales, HR, and IT networks after days of unsuccessful direct intrusion attempts. The breached materials reportedly included internal meeting transcripts, salary and job-offer information, and Slack messages discussing Anthropic's advanced artificial intelligence models.[3][4][6][8][9][11]
Anthropic responded that it cut off the unauthorized access the same day it occurred, confirming that no model weights, source code, customer prompts, or Claude responses were compromised. While the test was authorized under Anthropic's bug bounty program, the company noted Cathedral did not disclose all of its activities. Backed by Andreessen Horowitz and Sequoia Capital at a $1.4 billion valuation, Cathedral counts three federal agency contracts and co-founder Luke Farritor, who previously helped decode the ancient Herculaneum scrolls using machine learning.[6]
The revelation arrives as Anthropic publicly filed its S-1 registration statement for a planned initial public offering, reporting $19.7 billion in revenue over the prior 12 months. In the IPO filing, Anthropic disclosed previous AI security breaches alongside an intentional month-long pause in pre-training a new frontier base model, which was implemented over security concerns to bolster safeguards and prevent model weights from being stolen.[1][2][5][7][10]
Key facts
- Cathedral stated its AI agents gained access to Anthropic's sales, HR, and IT systems in under 12 hours via third-party software vulnerabilities.
- The accessed data included Anthropic meeting transcripts, salary and job offer details, and Slack messages concerning advanced models.
- Anthropic cut off Cathedral's access the same day and stated that model weights, source code, customer prompts, and Claude responses were not touched.
- Cathedral, backed by a16z and Sequoia at a $1.4 billion valuation, conducted the penetration test under Anthropic's bug bounty program, though Anthropic noted Cathedral failed to disclose all activities.
- Anthropic filed an S-1 registration statement for a planned IPO, revealing $19.7 billion in revenue over the trailing 12 months.
- Anthropic's IPO filing disclosed AI security breaches and a month-long pre-training pause for a frontier base model to improve security and protect model weights.
Sources · 9 sources
- TI
The Information@theinformationPost on X ·
Anthropic generated $19.7 billion in revenue over the past 12 months, according to its IPO filing. But another number caught Mostly Media founder @cjgustafson's attention. “They're doing over $6 million per employee. The median tech company does about half a million, 450K per employee.” Read more about the IPO filing: https://t.co/Y8VMxt2h1H
Open source - MT
MTS@MTSlivePost on X ·
SITUATION DETECTED: Anthropic recently delayed pre-training a new frontier base model for over a month to give itself time to ramp up security and prevent the model's weights from being stolen.
Open source - TE
Techmeme@TechmemePost on X ·
Cathedral, a months-old cybersecurity startup founded by ex-DOGE staffers, says it breached Anthropic's internal systems in under 12 hours using AI agents (Washington Post) (Visit Techmeme dot com for the link and full context!)
Open source - DE
DecryptArticle ·
AI Agents Found a Way Into Anthropic's Business Systems in Under 12 Hours Cathedral, a startup founded by ex-DOGE staffers, says its AI agents found paths into Anthropic's sales, HR, and IT systems through third-party software.
Open source - CB
Crypto BriefingArticle ·
Anthropic paused frontier model training over security fears, IPO filing shows Anthropic's security pause highlights the growing tension between rapid AI advancement and the need for robust safety measures, impacting investor confidence. The post Anthropic paused frontier model training over security fears, IPO filing shows appeared first on Crypto Briefing .
Open source - MT
MTS@MTSlivePost on X ·
SITUATION EXPLAINED: A months-old startup founded by ex-DOGE staffers used AI agents to break into Anthropic in under 12 hours. • Cathedral found paths to Anthropic's meeting transcripts, salary and job-offer details, and Slack messages about its advanced models • It got in through weaknesses in third-party business software, after days of failing to break into Anthropic directly • Anthropic says no customer prompts, Claude responses, source code, or model weights were touched, and it cut off access the same day • The test was authorized under Anthropic's bug bounty program, but Anthropic says Cathedral didn't disclose all of its activity • Cathedral is valued at $1.4B with backing from a16z and Sequoia, and 3 federal agencies already contract with it • Its co-founders include Luke Farritor, who helped decode the ancient Herculaneum scrolls with AI @theojaffee: "If a bunch of kids who have just started a company can do this, then potentially the Chinese aren't far behind. Aren't far behind? If anything, they're far ahead."
Open source - RE
Reuters@ReutersPost on X ·
Anthropic discloses month-long training pause, AI security breaches in IPO filing https://t.co/vrnjVxuV3O
Open source - PO
Polymarket@PolymarketPost on X ·
BREAKING: AI cybersecurity startup founded by ex-DOGE staffers breaches Anthropic's internal systems in under 12 hours using AI agents, accessing sensitive employee data & Slack messages.
Open source - TE
TechmemeArticle ·
Cathedral, a months-old cybersecurity startup founded by ex-DOGE staffers, says it breached Anthropic's internal systems in under 12 hours using AI agents (Washington Post) Washington Post : Cathedral, a months-old cybersecurity startup founded by ex-DOGE staffers, says it breached Anthropic's internal systems in under 12 hours using AI agents — The start-up firm Cathedral has told makers of the Claude chatbot that its researchers were able to quickly penetrate sensitive systems, a report reviewed by The Post shows.
Open source - RE
Reuters@ReutersPost on X ·
Anthropic discloses month-long training pause, AI security breaches in IPO filing https://t.co/81t8giJZ5K
Open source - TW
The Washington Post@washingtonpostPost on X ·
Exclusive: Cathedral, a months-old cybersecurity start-up, recently told Anthropic that its researchers had penetrated sensitive internal data and messaging systems at the AI company, according to a draft report reviewed by The Post. https://t.co/PY4AAyP9Pj
Open source