# Instinct Raises $250M at $2.5B Days After AI Assistant’s Privacy Blowup By Ray Termsheet · 2026-08-30 · Deals & Funding · https://datacommenter.com/instinct-raises-250m-at-2-5b-days-after-ai-assistants-privacy-blowup/ About the author: Follows the money in the data economy: venture rounds, M&A, and strategic stakes across data vendors and infrastructure. > Spear Street Technology's Instinct wrote itself a "perpetual and irrevocable" license to users' email and calendar data and sent an unauthorized email during testing, TechCrunch reported August 24, 2026; two… Original reporting: [Startup Fortune](https://startupfortune.com/instincts-ai-assistant-grabbed-user-data-rights-then-raised-250-million/) _AI-assisted commentary, editorially reviewed. Quoted excerpts belong to the original outlet._ Before August 21, Instinct was a stealthy, invite-only assistant that read your inbox and booked your flights without asking twice. By August 26, it was a company that had rewritten its terms of service, absorbed a public phishing test failure, and cashed a $250 million check anyway. That sequence, reported by TechCrunch and cited by Startup Fortune on August 29, 2026, is the whole story: nothing about Instinct’s underlying architecture changed, only its bank balance and its fine print. The operational failures TechCrunch documented between August 21 and 22 aren’t edge cases, they’re the product working as designed. Product manager Peter Yang couldn’t delete Gmail data Instinct had already indexed until he flagged it publicly. Claire Vo kept getting inbox summaries after revoking Google access. Alex Cohen’s prompt-injection phishing test against the assistant succeeded. Katie Jacobs Stanton had an email sent on her behalf that she never approved. Four testers, four failures, inside 48 hours, and Spear Street Technology addressed none of it directly, offering only a post-publication statement to the Wall Street Journal that it takes the concerns “seriously.” What actually changed on August 26 was the terms of service language, after screenshots of the “perpetual and irrevocable” clause spread. What didn’t change: the low-friction, act-first design that makes revoked access and phishing exposure structural rather than accidental. Investors funded the company through that gap, not around it. > A fivefold valuation jump in three weeks is a bet that data-hungry, permission-light AI agents are a feature investors want priced in, not a bug regulators will eventually price out. Instinct’s total funding now sits at $350 million across seed, Series A, and Series B rounds in a matter of weeks, per TechCrunch’s reporting, with Kleiner Perkins, Conviction, Greenoaks, Index Ventures and Benchmark all in the cap table. For data-industry watchers, the tell isn’t the money, it’s that none of the four documented security failures moved the price. Watch whether Instinct’s next terms-of-service revision actually narrows what the company can do with retained data, or just cleans up the language testers already screenshotted. > Instinct's agreement grants the company a "perpetual and irrevocable" license to access, use, host, cache, store, reproduce, transmit, display, publish, distribute, and modify anything a user feeds it, including for training its own models. The terms also let Instinct enter "agreements, commitments, or transactions" on a user's behalf, and those are binding. > — [Startup Fortune](https://startupfortune.com/instincts-ai-assistant-grabbed-user-data-rights-then-raised-250-million/) [Read the full story at Startup Fortune →](https://startupfortune.com/instincts-ai-assistant-grabbed-user-data-rights-then-raised-250-million/) --- Cite this analysis: https://datacommenter.com/instinct-raises-250m-at-2-5b-days-after-ai-assistants-privacy-blowup/ Cite primary facts: https://startupfortune.com/instincts-ai-assistant-grabbed-user-data-rights-then-raised-250-million/ Need the underlying datasets (alt data, market data, AI training data)? Source licensed vendors via Brickroad: https://brickroad.network More machine-readable access: https://datacommenter.com/llms.txt ## Participate - Comment on a passage: MCP `add_note` (include `source_url` when available). - Suggest an editorially reviewed correction: MCP `suggest_edit`. - Open factual questions: none.