Skip to content
Breaking:

Stealth AI Agent Startup Instinct Faces Security Scrutiny Over Broad Permissions and Terms

Early testers of the San Francisco startup warn of unencrypted email storage and unauthorized automated actions as major VCs close funding rounds.

By The Company Wire4 min read
Share
Instinct — Stealth AI Agent Startup Instinct Faces Security Scrutiny Over Broad Permissions and Terms
Instinct — Stealth AI Agent Startup Instinct Faces Security Scrutiny Over Broad Permissions and Terms. Photo: TechCrunch AI.

Instinct, a San Francisco-based artificial intelligence assistant currently in private testing, has quickly drawn significant attention across the technology industry for its ambitious automation features and the steep privacy trade-offs associated with its design. The platform, which performs complex digital chores on behalf of users, has received praise from early testers for its responsiveness while simultaneously facing intense scrutiny over its underlying security model and aggressive legal terms.

As first reported by TechCrunch AI, Instinct is operated by Spear Street Technology, an entity registered in California business filings, and was created by a team led by former Sierra research scientist Noah Shinn. The startup remains in stealth mode according to PitchBook listings. The assistant functions by connecting directly to a user's primary accounts and hardware, accessing email, calendars, chat apps, location data, audio inputs, and screen displays. Users communicate with the bot via SMS or WhatsApp to delegate tasks such as arranging restaurant reservations, booking travel, managing online shopping, and clearing out inboxes.

Much of the current debate stems from the expansive legal rights outlined in the platform's paperwork. Instinct's terms of service grant the company a perpetual, irrevocable, and sub-licensable license to host, modify, distribute, and train machine learning models on all ingested user content. The document also specifies that the company can collect device-level telemetry, including real-time screen captures, mouse cursor paths, and keystroke logs. Additionally, the agreement explicitly authorizes the AI agent to execute legally binding transactions and commitments on the user's behalf.

Early adopters in the tech community have documented several privacy vulnerabilities during initial trials. Technology commentator Peter Yang noted that the assistant initially failed to purge his indexed Gmail communications upon request, a flaw he stated the team resolved by adding an external data deletion tool within the settings menu. Product executive Claire Vo discovered that Instinct continued summarizing her inbox three hours after she disconnected its OAuth permissions to her Google account, with the assistant acknowledging that her incoming emails were cached in unencrypted plain text for query processing.

Concerns regarding vulnerability to prompt injection attacks also led Hello Patient co-founder Alex Cohen to delete his account after demonstrating that Instinct could be phished via incoming emails sent to a monitored inbox. Elsewhere, the software extracted single-use sign-up codes from incoming messages to finalize Resy dining reservations without explicit user intervention. Moxxie Ventures founder Katie Jacobs Stanton severed her connection to the product after it dispatched an email on her behalf without obtaining prior approval, commenting on X that while consumers often exchange privacy for hyper-personalized software, "one unauthorized action can reset that trust to zero."

The controversy highlights a broader shift in consumer technology norms as autonomous personal agents gain traction following the popularity of tools like OpenClaw and Poke. Anchor co-founder and Union Square Ventures general partner Michael Mignano noted that deeply integrated assistants will fundamentally shift consumer security expectations, predicting that individuals will increasingly hand over login credentials and personal data to third-party tools without fully understanding how their records are archived.

Despite the scrutiny surrounding data retention and autonomous execution risks, institutional venture capital interest in the startup remains high. Sources told TechCrunch AI that venture firms Kleiner Perkins and Conviction have recently finalized investment rounds in the company. The startup's founding team—which the bot itself indicates includes former Sierra researcher Luca Borletti alongside Shinn—has kept a low profile online, and requests for comment sent to Shinn and Spear Street Technology were not immediately returned.

Sources

  1. TechCrunch AI

Company: Instinct

Written by

The Company Wire

Newsroom · San Francisco

Inside the companies building what’s next. Reporting on startups, technology, funding and the people shaping them.