22 – 28 June 2026

On June 24, 2026, the company's services experienced elevated error rates. The issue affected claude.ai, the developer console, API, and tools like Claude Code and Claude Cowork.

The incident started at 13:16 UTC and was resolved by 14:56 UTC. This marks the third documented outage in the past eight days, following incidents on June 16 and 22.

All systems are currently operating normally. Anthropic has not disclosed details on the causes of the instability.

On June 22, 2026, Anthropic services encountered widespread errors, affecting the claude.ai web interface, API, Claude Code, and Claude Cowork.

The issue was detected at 19:14 UTC. Anthropic started investigating six minutes later and rolled out a fix by 19:20 UTC, with full recovery confirmed at 19:45 UTC.

This incident follows a similar outage on June 16, and the company continues to monitor system stability.

Claude Code v2.1.190 includes code changes suggesting Fable 5 is returning to standard subscriptions with a weekly usage cap.

The client now contains the string "You've used your Fable 5 usage for this week," and the previous "purchased separately from your plan" language has been removed. This marks a shift from the policy in place since June 23, where access to Fable 5 for Pro, Max, and Team users was suspended in favor of separately purchased credits.

Managing Director Chris Chauri previously mentioned that Mythos and Fable 5 would return in the coming days. Anthropic has not issued an official statement, and these changes are currently visible only within the client code.

According to Bloomberg, Anthropic and the Trump administration are approaching a deal to lift export controls on the Fable 5 and Mythos 5 models. Negotiations have been ongoing for several weeks and centered on system security.

Commerce Secretary Howard Lutnick is moving to address the concerns that prompted the initial restrictions. Sources indicate the controls could be lifted once administration officials provide their final consent.

The export controls were implemented in June 2026 following reports of a breach in the Mythos security architecture. President Trump later stated he no longer considers Anthropic a threat following a meeting with Dario Amodei at the G7 summit.

According to Bloomberg, Austria is negotiating with the EU to host Anthropic on its territory. The initiative follows US restrictions on foreign access to the company's most powerful models.

US export controls on the Mythos 5 and Fable 5 models have effectively cut off European users from Anthropic's advanced systems. Hosting the infrastructure in Austria could potentially provide the EU with access that bypasses these US restrictions.

The news received 9 points on Hacker News with two comments. Details regarding the hosting scheme and Anthropic's official stance remain undisclosed.

Users can now publish projects directly to Vercel from within Claude Design. After finishing a design, the system creates a new project in the connected Vercel account and returns a live link.

To enable the feature, users must add Vercel as a target in the Share menu and set up an MCP server to connect the services.

The integration allows publishing without leaving the Canvas workspace, cutting the time from design to live URL.

Anthropic has refreshed its technical documentation on system prompts in the API. Developers now have guidance on tuning model behavior through top-level instructions.

The materials cover methods for controlling tone, response formats, and safety limits. The documentation is in the official developer section on Anthropic's site.

Anthropic sent a letter to US senators and White House officials alleging that Alibaba Group ran a large-scale effort to access its models. According to Anthropic, thousands of fraudulent accounts were used to circumvent restrictions on Claude use in China.

The letter claims that operators linked to the Qwen AI lab deliberately tried to tap Claude's software development and agentic reasoning capabilities. Anthropic described this as the largest attempt by a Chinese company to exploit US tech.

Anthropic does not officially offer its products in China.

Security researchers report that the new Z.ai model from Zhipu AI, released this month, now matches top-tier American models in vulnerability detection. It has reached the performance level of Anthropic's Mythos in bug discovery tasks.

Z.ai still lags behind Anthropic and OpenAI products in other domains. However, cybersecurity was the specific area where American models held a clear lead, as previously demonstrated by Mythos testing in NSA systems.

According to Bloomberg, this parity creates pressure on the White House during its review of U.S. AI policy, including export restrictions on the Mythos 5 and Fable 5 models.