Product1 publisher3 min readPublished
Khosla's trust case for Wajo rests on safeguards still partly on the roadmap
Vinod Khosla is backing Wajo, a personal-agent startup claiming 10x growth since its late-September launch, on the view that trust decides the category. Wajo's own list of safeguards mixes features that work today with ones it has only announced, so the trust case has to be checked one feature at a time.
The Product Desk · Product desk
Drafted by a language model from the sources cited here and checked against its claim ledger before publication. How we use AISend a correction
What happened
- Meta's Muse, OpenAI's Dot and Instinct lead a personal-agent field that gains new competitors every week, according to TechCrunch.
- Wajo's Fo agent works inside iMessage and WhatsApp and can phone businesses or people to book or cancel services for its user.
- Founder Shivani Poddar held engineering roles at Meta, Google and DeepMind and worked on the first Facebook assistant.
- Jeff Dean and Gokul Rajaram have also backed Wajo, which did not say how much money it has raised.
Compiled by The Product DeskSomething wrong?How this is made
Why it matters
- decision If agents cannot unlearn, as Poddar says, the trust decision that counts is the scope of access granted at sign-up, and it falls to whoever connects the accounts before any safeguard has been tested.
- exposure People Fo phones take on the risk of a confused call without having signed up for anything, and the list meant to limit who can be called is announced but not yet shipped.
- constraint As quoted, Khosla's ad-business objection is aimed at Meta, so a buyer weighing Wajo against OpenAI's Dot has to compare the two on safeguards that actually ship.
A TechCrunch reporter asked Wajo's Fo agent to phone the reporter's partner with a reminder to feed the cats. The call did not go smoothly, and the agent's disclosures were less clear, the reporter wrote [8]. An earlier errand, calling an airline to check for upgrades, worked [7].
Those two calls are what people will actually ask a personal agent to do. Khosla's case for Wajo is about design choices a user does not see on either call. "I think the most important thing in an agent, because it's acting for you, is trust," he told TechCrunch, calling Wajo "the most trustworthy because it is architected to focus on trust and safety first, and some of the others are not" [3].
TechCrunch found that Fo works very much like Muse or Instinct [5]. With the features that close, Khosla's other argument is about who owns the agent. "If I give my data to Facebook, I'm the product, not the product I am buying, because they are in the ad business," he said [4]. His quoted remarks do not mention OpenAI or its Dot agent.
Poddar describes the risk in terms of time. "Trust is something you have to build from the get-go because one of the technological things that has not been solved yet is teaching agents to unlearn," she said. "If they get access to your personal data on day zero, they will always have it for the future" [13].
Wajo's safeguards split into what works now and what is promised. Fo can pay with a virtual card that keeps the user's details from the site [9], and Wajo has applied for a patent on that system [10]. Anyone who gets a call from Fo can tell it not to call again, and Wajo records the request [17]. The promised items are a "trusted contact" list for calls [16] and work in the coming months on how Fo behaves among several users [14]. Wajo also says its architecture follows Amazon's terms of service; Amazon has blocked tools like Muse [18].
On usage, Wajo offers a country count, 107, alongside its growth multiple [19]. Neither figure shows whether someone who booked one errand came back for a second.
Two questions sort these claims for anyone deciding whether to connect their own accounts or approve an agent for a team. The first is whether a safeguard is live today or only announced. The second is who carries the risk if it fails: the person who signed up, or someone the agent contacts who never did. On that grid the virtual card sits in the live, signed-up corner [9]. The trusted contact list sits in the announced, third-party corner [16], and the cat call happened there [8].
I'd switch on payments and calls to businesses now and keep Fo off calls to people until the contact list ships. The cost is the reminders and check-ins involving family and friends, the people who never chose the product. Instinct and Muse are also rolling out calling slowly, according to TechCrunch [15], so the same two questions apply to both agents Khosla measured Wajo against.
What to watch
- Whether Wajo ships the trusted contact list, and whether Fo's disclosure on personal calls gets clearer once it does.
- Whether Amazon treats Fo the way it treated Muse; a block would test Wajo's claim that its architecture follows Amazon's terms of service.
- Whether Wajo publishes user counts or repeat-use figures to sit behind its 107-country and 10x growth claims.