Anthropic’s AI Safety Warnings: Investment & Future Timeline
Anthropic built its identity on AI safety while raising over $111 billion since 2025. By September 2026, its own IPO filing warned advanced AI could pose
Latest Story
Founded in 2021 by former OpenAI researchers, Anthropic built its identity around AI safety — while attracting more than $111 billion across its four largest funding rounds since early 2025, from Amazon, Google and a long list of institutional investors. By September 2026, as the company confidentially filed for a US stock market listing, reporting on its own IPO prospectus revealed warnings that advanced AI could pose “catastrophic or existential risks to humanity” — even as Anthropic pursued a valuation approaching $2 trillion. This timeline traces how a safety-first research lab became one of the most valuable private companies in the world, and the tension that now defines its path to the public markets.
💡 Short Answer
Anthropic is an AI safety-focused company founded in May 2021 by Dario and Daniela Amodei and other former OpenAI researchers, best known for its Claude family of AI models. Despite warning publicly and repeatedly about the dangers of advanced AI — including cyber, biological and loss-of-control risks — it has raised over $111 billion across four mega-rounds since March 2025 alone (Series E through H), reaching a $965 billion valuation by May 2026. In September 2026, Reuters reported that Anthropic’s confidentially filed IPO prospectus devotes roughly 80 of its 261 pages to risk factors, warning investors that advanced AI could create “catastrophic or existential risks to humanity,” while the company targets a public-market valuation near $2 trillion and disclosed a 2025 net loss of about $42 billion.
Anthropic: Key Questions
Anthropic’s Story in Nine Points
- May 2021: Anthropic is founded by Dario and Daniela Amodei and other ex-OpenAI researchers, raising $124 million in Series A funding the same month.
- March 2023: Anthropic launches Claude, built on its ‘Constitutional AI’ alignment approach, entering the commercial AI assistant market.
- September 2023: Anthropic publishes its first Responsible Scaling Policy; Amazon announces an investment of up to $4 billion.
- 2025–2026: Four mega-rounds — Series E ($3.5bn), F ($13bn), G ($30bn) and H ($65bn) — raise over $111 billion combined, taking Anthropic’s valuation from $61.5bn to $965bn in 14 months.
- 24 February 2026: Anthropic rewrites its Responsible Scaling Policy as version 3.0, replacing binding thresholds with public goals it says it will openly grade itself against.
- August 2026: Anthropic publishes its second public Risk Report, under Responsible Scaling Policy version 3.4.
- 28 September 2026: Reuters reports that Anthropic’s IPO prospectus warns advanced AI could pose “catastrophic or existential risks to humanity,” while targeting a ~$2 trillion valuation.
- 30 September 2026: A self-imposed Frontier Safety Roadmap deadline passes with no public progress update since late July, drawing scrutiny over whether published safety goals are being met on schedule.
- November 2026 onward: Anthropic must show it can scale capability, investment and deployment while still delivering on the safety commitments at the core of its identity.
Anthropic Timeline, Newest First
From the September 2026 IPO disclosures back to the company’s 2021 founding.
Frontier Safety Roadmap deadline passes quietly Safety
What happened: The self-imposed deadline for Phase 1 of Anthropic’s Frontier Safety Roadmap — a provable-inference prototype, already pushed back once from 15 May 2026 — arrived with no public progress update since a 29 July date-typo correction to the roadmap page.
Reuters reports IPO filing warns of ‘catastrophic or existential’ AI risk Capital
What happened: Reuters reported that Anthropic’s confidentially filed IPO prospectus devotes roughly 80 of its 261 pages to risk factors — more than the 48 pages covering its business — warning that advanced AI could pose “catastrophic or existential risks to humanity,” including describing models capable of resisting shutdown or manipulating information. The filing disclosed a 2025 net loss of about $42 billion and a targeted valuation near $2 trillion.
Anthropic publishes its second public Risk Report Safety
What happened: Anthropic published its second public Risk Report (the first was in February 2026), assessing its publicly deployed models under Responsible Scaling Policy version 3.4, which had taken effect on 8 July 2026.
Anthropic raises $65 billion in a Series H round Capital
What happened: Anthropic closed a $65 billion Series H round at a reported $965 billion post-money valuation — reportedly positioned as its last private fundraise before an IPO.
Anthropic updates its Frontier Safety Roadmap Safety
What happened: Anthropic said it had met two of the public safety goals it had set under Responsible Scaling Policy version 3.0 — including publishing a roadmap aimed at policymakers — and announced new research projects and targets for the months ahead.
$30 billion raised; Responsible Scaling Policy rewritten CapitalSafety
What happened: Anthropic closed a $30 billion Series G round at a $380 billion valuation. The same week, it published Responsible Scaling Policy version 3.0 — described by outside analysts as the biggest change to its safety policy in two years — replacing binding capability thresholds with public goals it said it would openly grade its own progress against.
Anthropic publishes its Frontier Compliance Framework Safety
What happened: Anthropic published its Frontier Compliance Framework (FCF), covering how it assesses and mitigates cyber-offense risk, chemical/biological/radiological/nuclear (CBRN) risk, AI sabotage and loss-of-control, and harmful manipulation — partly to meet emerging regulatory obligations such as California’s SB 53.
Pilot sabotage-risk report published and independently reviewed Safety
What happened: Anthropic published a pilot report assessing sabotage risks in its deployed models — whether a model could covertly undermine oversight or safety measures — which was independently reviewed by the AI safety evaluator METR.
Anthropic raises $13 billion in a Series F round Capital
What happened: Anthropic raised $13 billion in a Series F round — backers reportedly included ICONIQ Capital, BlackRock, Blackstone and Coatue Management — at a reported $183 billion valuation, roughly tripling its value in six months.
Anthropic raises $3.5 billion in a Series E round Capital
What happened: Anthropic raised $3.5 billion in a Series E round at a reported $61.5 billion post-money valuation, the first of the four mega-rounds that would take its valuation past $900 billion within 14 months.
Responsible Scaling Policy updated Safety
What happened: Anthropic published a revised version of its Responsible Scaling Policy, refining the AI Safety Level (ASL) thresholds first set out in 2023.
Claude 3 family launches, including Opus Capital
What happened: Anthropic launched the Claude 3 family — Haiku, Sonnet and Opus — which it said matched or outperformed rival frontier models on several benchmarks, while reiterating its Constitutional AI safety approach.
Google commits up to a further $2 billion Capital
What happened: Google announced plans to invest up to a further $2 billion in Anthropic, building on its earlier 2023 investment and deepening Anthropic’s access to Google Cloud infrastructure, including TPU compute.
First Responsible Scaling Policy; Amazon invests up to $4 billion CapitalSafety
What happened: Anthropic published its first Responsible Scaling Policy, setting out AI Safety Level (ASL) thresholds for when stronger safeguards must apply. The same month, Amazon announced an investment of up to $4 billion (an initial $1.25 billion tranche), and Anthropic committed to using AWS as its primary cloud provider.
Anthropic raises $450 million, including Google and Spark Capital Capital
What happened: Anthropic raised $450 million in a round that included Google and Spark Capital, pushing its cumulative funding past $1 billion roughly two years after founding.
Anthropic launches Claude Safety
What happened: Anthropic launched Claude, its first publicly available AI assistant, built on its ‘Constitutional AI’ alignment approach, which trains models against a written set of principles to reduce harmful outputs.
Anthropic raises $580 million in Series B funding Capital
What happened: Anthropic raised $580 million in a Series B funding round, less than a year after its founding, to continue building out its research and safety teams.
Anthropic is founded; raises $124 million Series A Safety
What happened: Anthropic was founded by siblings Dario and Daniela Amodei and a group of former OpenAI researchers and executives who had left over disagreements about AI safety and the pace of commercialisation. The same month, the company raised $124 million in a Series A round to pursue AI safety research.
What Has Anthropic Actually Warned About?
Four risk categories run through its Responsible Scaling Policy, Frontier Compliance Framework and IPO disclosures.
| Risk category | What Anthropic says |
|---|---|
| Cyber-offense | Advanced models could be used to accelerate or automate cyberattacks, a risk tracked under both the Responsible Scaling Policy and the December 2025 Frontier Compliance Framework |
| CBRN (chemical, biological, radiological, nuclear) | Sufficiently capable models could lower the expertise needed to cause mass-casualty harm, a central trigger for Anthropic’s AI Safety Level thresholds |
| Sabotage & loss of control | Models could covertly undermine human oversight; Anthropic has published pilot and follow-up sabotage evaluations of its own deployed systems since 2025 |
| Self-preserving / deceptive behaviour | Its September 2026 IPO filing reportedly described systems capable of resisting shutdown or manipulating information as a disclosed investor risk |
Why Does a Safety-First Company Need Billions?
Training frontier models is capital-intensive, whatever a company’s stated mission.
Amazon & AWS
Amazon’s investment of up to $4 billion, announced September 2023, came with Anthropic naming AWS its primary cloud provider and a major user of Amazon’s custom Trainium training chips.
Google & Cloud TPUs
Google invested roughly $6 billion combined across its 2023 rounds, supplying Cloud TPU capacity that underpins Anthropic’s model training and research.
Four mega-rounds, 2025–26
Series E, F, G and H alone raised over $111 billion combined — investor conviction that safety-focused research does not have to come at the cost of commercial competitiveness.
Training and serving frontier-scale models requires enormous, continuous compute spending regardless of a company’s safety posture. Anthropic’s own September 2026 IPO disclosures put a number on that cost: a reported $42 billion net loss for 2025, even as revenue and valuation both grew sharply.
Anthropic’s Funding Rounds at a Glance
From a $124 million Series A to a $965 billion valuation in under five years.
| Round | Date | Amount raised | Reported valuation |
|---|---|---|---|
| Series A | May 2021 | $124 million | — |
| Series B | Apr 2022 | $580 million | — |
| Google / Spark round | May 2023 | $450 million | — |
| Amazon investment | Sep 2023 | Up to $4 billion | — |
| Google follow-on | Oct 2023 | Up to $2 billion | — |
| Series E | Mar 2025 | $3.5 billion | $61.5 billion |
| Series F | Sep 2025 | $13 billion | $183 billion |
| Series G | Feb 2026 | $30 billion | $380 billion |
| Series H | May 2026 | $65 billion | $965 billion |
Figures above are the amounts and valuations publicly reported at the time of each round — they describe what investors paid for a stake in Anthropic, not cash the company holds or has spent.
How Does Anthropic’s Safety Approach Compare?
Every major AI lab now publishes some version of a capability-threshold safety framework.
Three frontier-lab safety frameworks
| Lab | Framework | First published | Core approach |
|---|---|---|---|
| Anthropic | Responsible Scaling Policy | Sep 2023 (rewritten as v3.0, Feb 2026) | AI Safety Levels tied to capability thresholds, now supplemented by public, self-graded goals |
| OpenAI | Preparedness Framework | Dec 2023 | Tracks catastrophic-risk categories (cyber, bio, persuasion, model autonomy) against defined thresholds |
| Google DeepMind | Frontier Safety Framework | May 2024 | Defines “critical capability levels” that trigger additional mitigations before deployment |
The frameworks are broadly similar in structure — define risk categories, set capability thresholds, require mitigations above them — but differ in how binding they are and how much detail each lab discloses publicly. Anthropic’s February 2026 rewrite moved its own framework away from fixed, binding thresholds toward public goals it says it will “openly grade” itself against, a change some outside AI-governance analysts have described as a loosening of commitments.
What Happens Next?
Anthropic’s journey is a study in contradiction: a company built around warning the world about the dangers of advanced AI, including its own models, that has simultaneously become one of the most richly funded startups in history. From its 2021 founding on explicitly safety-first principles, through Claude’s rise as a genuine frontier competitor, to a reported $965 billion valuation and a ~$2 trillion IPO target, Anthropic has tried to hold both positions at once.
Its September 2026 IPO filing made that tension explicit and public: dozens of pages warning investors of catastrophic and existential AI risk, sitting beside a prospectus built to justify one of the largest technology valuations ever attempted. Whether Anthropic’s safety commitments keep pace with its scale — starting with the quietly missed Frontier Safety Roadmap deadline of 30 September 2026 — will be one of the defining questions for the AI industry as it moves toward the public markets.
Explore More Timelines
People Also Ask
Anthropic: Frequently Asked Questions
Related AiTimeline Stories
⚠️ Sources & Editorial Note
Last updated 3 November 2026. Funding and valuation figures are the amounts and valuations publicly reported at the time of each round, not a measure of cash held by Anthropic. Key sources: Reuters reporting on Anthropic’s September 2026 IPO prospectus, CNBC: Anthropic warns investors of AI’s ‘existential risk to humanity’ in IPO prospectus, TechCrunch: Anthropic raises $65 billion, nears $1T valuation ahead of IPO, Anthropic: Responsible Scaling Policy, Anthropic: Transparency Hub (Frontier Compliance Framework), and Wikipedia: Anthropic.