Claude Code users complain their chat records are being mysteriously wiped out

Claude Code users report that conversation transcripts are silently deleted via the default cleanupPeriodDays=30 setting, wiping older .jsonl files on each app startup. Anthropic says the 30-day retention policy is a security-by-design measure to limit exposure to sensitive data (e.g., code/credentials) but critics highlight missing disclosure and no recovery/soft-delete option. Some reports suggest changing retention can fail due to deletion being keyed to file modification time (mtime) rather than real last activity, prompting calls for product-level transparency and fixes.

Analysis

This is less a product glitch than a trust-tax on AI coding workflows. The market mechanism is procurement friction: once buyers realize working context can disappear without an audit trail, enterprise expansion slows until vendors prove retention controls, restore paths, and explicit admin disclosure. That shifts bargaining power toward incumbents with governance layers and away from standalone copilots that monetize speed but not durability.

Near term, the selloff risk is mostly reputational, not revenue-driven, but the second-order beneficiaries are privacy, backup, DLP, and e-discovery vendors. Over 1-3 months, expect IT/security teams to tighten policies around local storage and transcript handling, which should modestly help names with compliance-heavy positioning while pressuring AI workflow tools that rely on developer trust. The real downside catalyst would be evidence of broader silent-loss incidents or customer complaints in regulated verticals; that would turn a UX issue into a sales-cycle issue.

Contrarian view: the default retention policy itself is probably defensible, and the consensus may be overpricing the headline while underpricing the disclosure failure. If the vendor quickly ships visible warnings and a recoverable delete flow, this becomes a short-lived community issue rather than a demand shock. Falsify the bear case if enterprise adoption, seat counts, or developer engagement stay intact over the next two quarters despite the noise.

Key Decisions for Investors

• Watchlist, not trade, on CGAC if it is a cyber/privacy beneficiary; buy on weakness only if management later cites stronger demand for retention, audit, or DLP features tied to AI workflows.
• If TSTS is exposed to AI developer tooling, fade rallies over the next 1-3 weeks; use a small short only if follow-on issues show churn or procurement delays, with a stop on any rapid product fix and customer reassurance.
• Long PANW or CRWD on a 1-3 month horizon as a relative winner from higher enterprise demand for governance, logging, and data-loss controls; risk/reward is better than owning unproven copilots because the revenue model is already tied to security budgets.
• Pair trade: long MSFT / short WCLD for 1-3 months. MSFT is better insulated because it can bundle workflow, identity, and compliance controls, while higher-beta software multiples remain vulnerable if buyers start demanding durable context storage before expanding AI seats.
• Set a catalyst alert: if the vendor announces soft-delete, restore, or explicit first-run disclosure, cover any short exposure immediately; if there is no measurable slowdown in enterprise AI adoption by the next earnings cycle, this should be treated as noise.