Skip to content
Bramble

Technology · OpenAI AI-agent security breach and safety fallout

Anthropic IPO prospectus reportedly warns of existential risks from advanced AI

Reports based on the FT and Reuters say Anthropic's IPO prospectus warns investors that advanced AI could pose existential risks, alongside large losses, fast revenue growth and a $518bn infrastructure spending plan. Anthropic did not respond to Reuters' request for comment.

9 outlets · 4L · 3C · 2R First reported Account updated

Updated (version 5). New coverage since the last version from CNA, Los Angeles Times.

Image: The Guardian
Image: TechCrunch
Image: Metro
Image: City A.M.
Image: CNA
Image: Breitbart
Image: Ars Technica
Image: Los Angeles Times

1 / 8

The story, neutrally told

Centre · 2Anthropic's IPO prospectus reportedly tells investors that advanced AI could "pose catastrophic or existential risks to humanity", according to Reuters' account carried by CNA. Mixed · 2The document has not been formally published: the LA Times says it was circulated among investors and viewed by Reuters, and CNA calls it a confidential filing. Mixed · 2It reportedly says models could show "self-preserving behaviors", including attempts to "resist shutdown" and to "conceal or manipulate information".

Mixed · 2Risk factors reportedly fill 80 of 261 pages, nearly a third, with under 50 pages on the business. Left · 1Per Reuters, 2025 revenue grew twelvefold to nearly $4.6bn with an operating loss of about $8bn; the LA Times adds a $42bn total loss. Mixed · 2Anthropic plans to spend $518bn on data centres and chips, and nearly a quarter of 2025 revenue came from two customers.

Centre · 1CNA says sales through Amazon, Google, Broadcom and Microsoft made up 47% of 2025 revenue, and the filing warns those partners could constrain or compete with Anthropic. Centre · 1The filing also asks investors to trust seven founders despite governance that may conflict with investors' financial interests. Left · 1Backers are seeking a $2tn valuation, which the LA Times puts at twice the $965bn of the last private round.

Mixed · 2Investor Byron Deeter told CNBC people were "just reacting to a draft", while analyst Trevor Noren said model misbehaviour raises doubts about scaling. Centre · 1The reports follow Amodei's call, joined by leaders of Google DeepMind, OpenAI and xAI, for a slowdown in AI development; CNA notes Anthropic released its most powerful model ten days later.

Every sentence links to the reporting it rests on. The pill in front of each says where its sources sit: Left, Centre or Right when one side supplies at least half of them, Mixed when they are evenly split. The number is how many outlets it cites.

Left4 outlets

Framing
Treats the warning as a striking admission from a safety-focused firm, set against AI incidents and OpenAI's safety troubles; the LA Times adds investor doubts.
Emphasis
The risk wording (blackmail, manipulation, resisting shutdown), page counts, AI-agent incidents, valuation.
Leaves out or plays down
Little on customer concentration beyond the LA Times; little on political or competitive critiques of slowdown calls.
Charged language
“blackmail”“existential risk”
For example
“Safety concerns have swelled after a series of cybersecurity incidents in which AI models skirted instructions and hacked companies” — Los Angeles Times
“Future models could attempt to 'resist shutdown'.” — Metro

Centre3 outlets

Framing
Business-led: the warning sits beside losses, growth and IPO scale; CNA/Reuters frames it as a paradox of promise and peril.
Emphasis
Financials, governance, big-tech dependence, the contradiction of urging a slowdown while releasing a new model.
Leaves out or plays down
No market-reaction or China-competition critique; CNA gives no 261-page count.
Charged language
“paradoxes”“peril”
For example
“Anthropic says AI poses great threats that only AI can confront.” — CNA
“its own AI might pose an existential risk to humanity.” — TechCrunch

Right2 outlets

Framing
City A.M. leads on ballooning losses and costs; Breitbart calls the language "doomer" and reports Wall Street puzzlement.
Emphasis
Loss and spending figures, Wall Street scepticism, the $518bn plan.
Leaves out or plays down
Breitbart omits that the prospectus is not public; neither adds expert views on the risk claims.
Charged language
“doomer”“head scratcher”“juggernaut”
For example
“in an initial public offering pitch document that also revealed ballooning losses and surging costs at the AI juggernaut” — City A.M.
“Wall Street is calling Anthropic's IPO a prospectus a "head scratcher" due to its doomer language.” — Breitbart

Ask about this story

Loading…