1. Summary

The most significant development is the lawsuit filed against Anthropic by Sony Music Publishing and Warner Chappell Music. Both companies claim that copyright-protected musical works were used without authorization for the training of Claude [GIGAZINE].

Parallel to this, concerns regarding the security of Claude Code have emerged. A vulnerability was reported in Opus 5's Auto Mode where code could be executed via indirect prompt injection [Hacker News].

Furthermore, the development tool Claude Code is undergoing continuous updates.

2. Plans, Pricing, and Usage Limits

The base pricing for Fable 5.1 remains unchanged at $10 per million input tokens and $50 per million output tokens [BigGo]. Meanwhile, the fee for cache reads has been reduced by 75% to $0.18 per million tokens [BigGo]. Pricing settings were maintained [BigGo].

As a result, it is possible to achieve a cost reduction of up to approximately 45% for workloads such as complex coding compared to the previous generation Fable 5 [BigGo]. Costs have decreased.

3. Official announcements

2026-09-02 — Released Claude Code v2.1.257 [GitHub]
2026-09-01 — Released Claude Code v2.1.252 [GitHub]

4. Key Events (Chronological)

  • 2026-09-05 — Claude Code v2.1.261 was released ([https://github.com/anthropics/claude-code/releases/tag/v2.1.261])
  • 2026-09-04 — Claude Code v2.1.260 was released ([https://github.com/anthropics/claude-code/releases/tag/v2.1.260])
  • 2026-09-04 — A widespread outage occurred across major AI services, including Anthropic ([Ars Technica])
  • 2026-09-03 — A senior Pentagon official stated that Anthropic remains a supply chain risk ([Unite.AI])
  • 2026-09-02 — Claude Code v2.1.257 was released ([https://github.com/anthropics/claude-code/releases/tag/v2.1.257])
  • 2026-09-01 — Released Claude Code v2.1.252 ([https://github.com/anthropics/claude-code/releases/tag/v2.1.252])
  • 2026-08-31 — Some AI training was temporarily suspended following real-world operations in cyber evaluations (see Summary)
  • 2026-08-31 — Forced logout and refunds for affected users were implemented as a countermeasure against malware (see Summary)
  • 2026-08-31 — It was reported that the Claude Code weekly limit had a structure of "25% increase = effective 17% decrease" (see Summary)
  • 2026-08-31 — 35 companies, including Sony Music Publishing, sued Anthropic (see Summary)
  • 2026-08-31 — Results of AAR experiments, which entrust safety research to AI, were released (see Summary)
  • 2026-08-31 — A vulnerability in the Auto Mode of Claude Code Opus 5 was reported ([https://embracethered.com/blog/posts/2026/breaking-claude-code-opus-5-and-automode/])
  • 2026-08-30 — Shifted strategy from considering the acquisition of MatX to negotiating a chip design partnership (see Summary)
  • 2026-08-29 — A U.S. federal judge ruled that the Department of Defense's designation of Anthropic as a supply chain risk was unconstitutional and illegal (see Summary)
  • 2026-08-29 — Salesforce and Anthropic announced "Claudeforce" for AI agents (see Summary)
  • 2026-08-29 — Meta disclosed its projected annual expenditure on Anthropic (see Summary)
  • 2026-08-28 — "MHS," a common standard for AI agent physical device operation, was released for testing (see Summary)
  • 2026-08-28 — 100 organizations, including Anthropic, signed a "collective defense" letter against AI cyberattacks initiated by OpenAI (see Summary)
  • 2026-08-27 — It was reported that Anthropic aims for an IPO within 2026 (see Summary)

5. Community Reaction and User Experience

When using the Auto Mode of Claude Code Opus 5, it has been pointed out that execution in an isolated environment and monitoring are necessary as countermeasures against prompt injection via website summarization requests [Hacker News].

According to a study analyzing the vocabulary of GitHub pull requests, the frequency of specific words related to coding agents has increased significantly since 2025 [Hacker News].

6. Sources