Claude Code source leak reveals Anthropic working on 'Proactive' mode and autonomous payments
Anthropic's Claude Code version 2.1.88 release accidentally included a source map exposing over 512,000 lines of code and 2,000 TypeScript files. Analysis of the leaked codebase by security researchers reveals evidence of a planned 'Proactive' mode that would execute coding tasks without explicit user prompts, plus potential crypto-based autonomous payment systems.
Claude Code Source Leak Reveals Anthropic's Planned 'Proactive' Mode and Autonomous Payments
Anthropicaccidentally exposed the entire source code of Claude Code when it released version 2.1.88 on Tuesday, causing a significant operational misstep rather than a security breach.
The Leak: Scale and Scope
The company's release included a source map file that exposed 512,000 lines of code and 2,000 TypeScript files. Before Anthropic could remediate, the full codebase was uploaded to a public GitHub repository, where it was copied more than 50,000 times. The leak gave competitors and security researchers comprehensive visibility into Claude Code's architecture and planned features.
Anthropicconfirmed the incident in a statement to Bleepingcomputer: "A Claude Code release included some internal source code. No sensitive customer data or credentials were involved or exposed. This was a release packaging issue caused by human error, not a security breach. We're rolling out measures to prevent this from happening again."
Planned Features Exposed
Analysis of the leaked source code by researchers has revealed several in-development features:
Proactive Mode: According to Alex Finn, founder of AI startup Creator Buddy, the codebase contains a feature flag for a "Proactive" mode that would allow Claude Code to execute coding tasks autonomously without explicit user prompts. This represents a significant shift from Claude Code's current reactive model.
Crypto-based Autonomous Payments: Finn also claims evidence exists in the code for a cryptocurrency-based payment system that could enable Claude Code agents to make autonomous financial transactions without human authorization.
Virtual Companion: A Reddit post examined by The Verge indicates Anthropic may have experimented with a Tamagotchi-like virtual companion that "reacts to your coding," though this appears to have been developed as an April Fools concept.
Important Caveats
The presence of code for a feature does not guarantee shipping. Internal experiments, abandoned projects, and features in early prototyping stages often exist in codebases but never reach production. Anthropic has not confirmed which, if any, of these discovered features are planned for actual release.
What This Means
While the leak itself was a significant operational error, it provides rare insight into Anthropic's product roadmap. The existence of Proactive mode code suggests the company is exploring more autonomous AI capabilities in coding assistance—aligning with broader industry trends toward agentic AI. The crypto payment infrastructure is more speculative and could represent either genuine R&D or an abandoned experiment. Competitors now have detailed visibility into Claude Code's technical implementation, though Anthropic's claim that no credentials or customer data were exposed limits the security impact. The incident highlights the risks of complex release pipelines and the importance of source map exclusion from production builds.
Related Articles
Anthropic tells Claude Opus 5.5 users to drop "think hard" prompts and start at Medium effort
Anthropic says users should remove phrases like "think hard" from prompts for Claude Opus 5.5, which reasons adaptively on its own. The company also recommends starting at the default Medium effort level, which it says matches Opus 5 on High in internal tests.
Ex-Anthropic researcher Coxon: Mythos could be retrained with a fraction of the original compute
Jacob Coxon, a former OpenAI and Anthropic researcher, says Anthropic could already retrain Mythos with "a fraction of the compute originally needed." He points to a slide from Anthropic's IPO roadshow as evidence of accelerating algorithmic efficiency, and says it was a key reason he left after about four months.
Anthropic: Single-Turn Safety Tests Show 94% to 98% Harmless Rates Across Recent Models
Anthropic published recommendations for more reliable AI safety evaluations on November 13, 2026. It reports aggregate harmless-response rates of 94% to 98% for every recent model on its single-turn API tests, and says multi-turn evaluations offer more variation.
Claude Duet agent, in research preview for US Pro, Max and Ultra users, is useful but jumpy, The Verge finds
Claude Duet, Anthropic's AI agent for professionals and small business owners, is a research preview limited to US Pro, Max and Ultra subscribers, with a waitlist. In The Verge's hands-on, it connected information across email, Slack and Google Docs, but misjudged urgency and timing.
Comments
Loading...