• Anthropic confirmed elevated error rates across its newest models—Fable 5.1, Mythos 5.1, and Opus 5—that began at 00:57 UTC and were resolved by 02:35 UTC, affecting over 1,600 users.
  • The incident, though brief, underscores the operational risks as AI services become embedded in enterprise workflows, and comes as Anthropic prepares for a public offering amid rapid revenue growth and deepening losses.
  • Despite the disruption, Anthropic's official status page now lists all services operational, with no ongoing incident reported.

Anthropic's Claude appears to be operating normally again after a service disruption earlier today, according to the company's status page. The incident, which began at 00:57 UTC with elevated error rates across the Fable 5.1, Mythos 5.1, and Opus 5 models, was resolved by 02:35 UTC after engineers identified the cause and normalized success rates. Crowdsourced reports from Downdetector indicated more than 1,600 users encountered issues, including access failures, API errors, and "overloaded" responses.

The disruption, while short-lived, affected the company's most advanced models—including those central to its newest offerings. Fable 5.1 and Mythos 5.1 were introduced on September 1 and are marketed as advanced models for coding and knowledge work, while Opus 5 forms part of Anthropic's flagship lineup. For developers and enterprises that have integrated these models into production pipelines, even an 80-minute outage can disrupt software deployments, customer support automation, and data analysis workflows.

This isn't the first hiccup for the San Francisco–based AI developer. A smaller incident on September 15 affected Mythos 5.1 and Fable 5.1, and a Google Play issue on September 16 impacted Android subscription actions. But the September 22 event stands out for its scope and timing, arriving just as Anthropic ramps up its product cadence. The company launched Opus 5.5 on the same day and followed with Sonnet 5.5 on September 28, which it says is 30% faster and up to 30% cheaper than its predecessor.

Operational Strain Meets Financial Reality

The outage highlights a broader challenge for Anthropic and its peers: as AI becomes critical infrastructure, reliability is emerging as a competitive battleground. Anthropic's disclosed compute and infrastructure spending reached $7.33 billion in 2025, a reflection of the enormous resources required to train and run frontier models. Yet the company remains deeply loss-making. Its 2025 revenue hit nearly $4.6 billion—a 12-fold year-over-year increase—but operating losses exceeded $8 billion, excluding certain financing-linked writedowns. Total operating expenses were about $13 billion.

A reported net loss of roughly $42 billion for 2025 includes a $34 billion accounting charge tied to convertible financing, according to IPO prospectus materials seen by Reuters. The company's future cloud and computing obligations total a staggering $518 billion, underscoring the capital intensity of the AI arms race.

"It's a reputational challenge at a sensitive moment," said a person familiar with the company's investor relations, speaking on condition of anonymity. "Anthropic is trying to demonstrate it can operate at global scale while preparing for a public listing."

Anthropic did not respond to a request for comment on the root cause of the outage. A company spokesperson previously stated that the issue was identified and fixed, but no technical details were disclosed.

Governance and IPO Plans Take Shape

As Anthropic moves toward an IPO, its governance structure is drawing attention. CEO Dario Amodei and co-founder Daniela Amodei, who serves as president and chair of the board, remain at the helm. Reuters reports that a "Founder LLC" composed of seven co-founders would control a Class F share representing 50.1% of voting power over key corporate matters. The company is retaining its public-benefit-corporation structure, which explicitly allows it to balance investor interests with its stated public-interest mission—a structure that could conflict with short-term financial performance.

Other senior appointments include CFO Krishna Rao (2024), CTO Rahul Patil (2025), Chief Commercial Officer Paul Smith (2025), and international managing director Chris Ciauri (2025).

The outage also feeds into a wider policy discussion. While not a safety incident, it raises questions about whether AI providers should face formal resilience and incident-reporting obligations as their services become embedded in essential commercial and public-sector workflows. Anthropic has advocated for transparency requirements and independent evaluation, and its Frontier Compliance Framework aims to address catastrophic risks. But operational reliability remains a separate, pressing concern.

For now, users can breathe a sigh of relief: all Claude services are listed as operational. But for enterprises building on the platform, the incident serves as a reminder to invest in redundancy—model-routing systems, fallback providers, and graceful degradation—before the next outage hits.

Correction: An earlier version of this article misstated the date of the September 15 incident. It affected Mythos 5.1 and Fable 5.1, not Opus 5.