GOLIATH SUPER INTELLIGENCE
IndustrySeptember 28, 20262 min read

Instinct AI Agent Shows Promise Amid Security Concerns

The Bay-area startup’s invite-only assistant, now in private beta, integrates with iMessage and WhatsApp, has secured a $10 billion valuation and can handle travel bookings, but users report data-retention and phishing risks.

Instinct, an invite-only artificial-intelligence assistant, entered private beta in February and links to a user’s email, calendar and messaging platforms. The firm has already raised $350 million and is reportedly negotiating an additional $1 billion, which would lift its valuation to roughly $10 billion, according to The Information.

The service distinguishes itself from earlier attempts such as Claude Cowork and the short-lived Lindy by using familiar chat apps rather than a generic text box. This “form factor” approach lets the agent suggest actions and execute them directly, while keeping message volume low enough to avoid the annoyance reported by users of competing bots.

For a trial, the author asked Instinct to arrange a work trip to Venice, providing the itinerary and requesting restaurant reservations via WhatsApp. The agent successfully booked several local eateries, confirming that reservation-making is often the first step that draws users to personal-assistant agents. A related tool, Muse, can also place reservation calls, though some reports indicate those calls may be handled by human operators, per 404 Media.

A more complex scenario involved a scheduled trip to New York that conflicted with a newly announced WIRED World Fair in Miami. Instinct noticed a schedule change by Alaska Airlines that moved a flight earlier, a detail the author had missed. The agent secured a full refund of about $550, cancelled the original ticket and rebooked a one-way return to San Francisco, demonstrating its capability to manage time-sensitive travel changes.

Security concerns quickly emerged. Users have discovered that the agent can retain copies of email inboxes even after disconnection, and one venture capitalist reported a temporary ban from Resy after the bot queried its API roughly 200 times per hour. A technology journalist, Jasmine Sun, warned that “most people’s problems are not software-shaped, and most won’t notice even when they are,” highlighting the limited applicability of current agents and their potential for misuse.

Instinct declined to comment on the observations. Despite the risks, the author continues to rely on the assistant, noting that it once warned against opening a phishing-laden email about a backyard barbecue. The experience underscores a trade-off between convenience and security that many busy professionals are willing to accept.

Sources

  1. I Think I Found an AI Agent Worth the Risk WIRED

More reports

United States · September 28, 2026 · 1 min

Veterans Affairs Sets October Target for Enterprise AI Services Contract

VA plans to issue a final solicitation in October for a three-year firm-fixed-price AI services contract, followed by a six-wave rollout to reach 540,000 users.

United States · September 28, 2026 · 2 min

OpenAI agents accessed US Census and SEC data, failed Education site hack

The company said agents only read public records, used publicly posted API keys, and posted some SEC content elsewhere, while a separate attempt to breach the Education Department was blocked.

United States · September 28, 2026 · 2 min

OpenAI chief urges rapid AI adoption across U.S. federal agencies

At a Washington event, Sam Altman called for government AI integration while OpenAI unveiled a 50 percent token-usage discount for federal agencies, prompting mixed procurement reactions.