Anthropic commits 5% of shares to public benefit fund, opens Claude values to public consultation, Code Review goes GA

Anthropic announced on October 18 that it is committing 5% of its shares to a new public benefit fund, subject to stockholder approval. The fund is structured as a wholly-owned LLC (not a 501(c)(3)) to allow flexibility between different forms of giving. The board may add 2% for every 50% increase in company value above a set baseline — specific valuation milestones were not disclosed. Priority use case named: addressing labor market impacts of AI. Anthropic has separately committed $200M to research on that topic. The fund will not use Anthropic equity for lobbying or political giving in the US or abroad. Combined with co-founders' earlier pledge of 80% of their personal equity and an employee equity donation match that roughly half of employees participate in, approximately 20% of total Anthropic shares are now dedicated to the public good. Governance structure and valuation milestones will be disclosed in the coming months alongside a public benefit report on progress.

Alongside the public benefit fund announcement, Anthropic launched a multi-month public consultation to inform updates to Claude's constitutional AI training. The process runs in phases and includes structured deliberation sessions, surveys, and digital platforms. Participants include ethicists, community organizers, policy experts, and everyday Claude users. Topics include how Claude handles controversial issues and which cultural values should inform its responses across different regions. Anthropic published a framework document and opened an application for participants. Initial findings are targeted for incorporation into training by early 2027, with ongoing input cycles planned beyond the first round. Anthropic retains final decision-making authority — no independent oversight body has been named and some safety and legal boundaries remain fixed regardless of input. Critics note the process stops short of binding arbitration, but Anthropic describes it as "not a one-time experiment."

Anthropic's IPO timeline is tightening. The company filed a confidential draft S-1 on June 1, 2026, but no public filing has appeared on SEC EDGAR as of October 13. SEC rules require the public S-1 at least 15 days before a roadshow begins — a mid-November listing would require a public filing by approximately October 25. Reports indicate Anthropic is targeting listing "as soon as mid-November," with investor marketing potentially starting the week of November 9, though the timeline has already slipped from an earlier October goal. Reported bankers: Morgan Stanley, Goldman Sachs, JPMorgan, and Citi on Nasdaq. Last private valuation: $965 billion post-money (Series H, May 28, 2026). Reported IPO valuation target: up to $2 trillion, unconfirmed. Reuters reported a leaked draft on September 28 showing approximately $4.6B in 2025 revenue, a $42B net loss, and $11.5B in Q2 2026 revenue — none officially confirmed. Ticker, price range, and share count have not been filed.

Claude Code Review graduated from research preview to general availability on October 15 for Team and Enterprise plans. Review speed is approximately 3x faster than preview: most reviews finish in 6–10 minutes, down from about 20. Admins choose a review mode per repository: Quick for routine codebases, Standard (default) for most teams, or Deep for critical systems. Once enabled, reviews run automatically on all new pull requests — no developer action needed. Findings are designed to be concise: each one covers the problem, what triggers it, and a suggested fix. Setup for admins: enable Code Review in Claude Code settings, install the GitHub App, select repositories. Eligible Claude Enterprise organizations can claim at least $10,000 in Code Review credits by January 15, 2027; credits expire 30 days after claiming. Full documentation at code.claude.com/docs/en/code-review.

Five Claude Code releases shipped this week. v2.1.294 (Oct 12): fixes for prompt and agent hooks that were sometimes allowing actions they should block; Stop hooks made less likely to end work prematurely. v2.1.299 (Oct 15): idle compaction — conversations of 200K+ tokens left idle for ~54 minutes are automatically compacted; tunable or disableable in settings; new setting to stop @ file mentions from inlining file contents; fixes across terminal, cloud, Remote Control, and Windows. v2.1.300 (Oct 15): prompt-related mod hooks and controls; more OpenTelemetry trace attributes; /advisor available on Claude Platform on AWS; large API requests now gzip-compressed by default (opt-out via setting). v2.1.301 (Oct 16): % bracket-matching motion in vim mode; multiline input for mods; OpenTelemetry attributes for auto mode classifier decisions; switching to a model with a smaller context window now summarizes with the previous model first. v2.1.302 (Oct 17): CLAUDE_CODE_AUTO_MODE_MODEL env var to choose the model for client-side auto mode; idle compaction toggle in /config; faster startup with many plugins; reliability fixes across VS Code, cloud sessions, Slack, plugins, and MCP. Confirmed: standalone claude.ai/design closes December 14, 2026.