Amazon Drops Data Center NDAs as Personal AI Agent Startups Bet on Privacy and Trust

Amazon has said it will stop using non-disclosure agreements when negotiating data center deals with local governments. This decision is not an isolated event—it closely follows a similar choice made by Microsoft earlier this year. Two hyperscalers successively abandoning the same negotiating tool suggests that secrecy practices around AI infrastructure are under growing external pressure.

The pressure comes from the community level. The secrecy practices of such agreements have fueled growing community backlash against AI infrastructure and have led to hundreds of proposed and enacted moratoriums from New York to San Francisco and beyond. In other words, NDAs are no longer just a technical detail of business negotiations—they have become a trigger point for local political resistance.

This shift was the central topic of the latest episode of the TechCrunch 'Equity' podcast. Hosts Anthony Ha, Sean O'Kane, and Rebecca Bellan explored whether abandoning NDAs will change how data center deals are made.

It is worth noting that the podcast raised an open question rather than offering a conclusion. Based on the available information, whether Amazon's abandonment of NDAs will truly change how data center deals are made remains uncertain. It can be inferred that, against the backdrop of community backlash and frequent moratoriums, the move is intended to ease political resistance at the local level—but the confidence in this inference is limited, because the source did not provide direct explanations from Amazon or Microsoft about their motivations.

The same podcast episode turned its attention to personal AI agent startups. The hosts discussed how Tab, Underdog, and Hark position privacy as their core product, and suggested that trust may ultimately be what personal AI startups are selling.

Behind this judgment is a practical premise: consumers need to be willing to let AI agents access their inboxes, files, and credit cards. From the podcast discussion, it can be inferred that these startups are using privacy and trust as differentiating selling points in exchange for users' authorization to access sensitive data. However, whether these companies' privacy-centered products can earn consumer trust remains to be verified.

The podcast also explored a business model that could turn personal agents into salespeople. Whether this model is viable is likewise not given a clear answer in the source.

For personal agents to be useful, they must be able to access external websites and services. The podcast focused on how United, Apple, and Amazon have set up barriers against external agents. This defensive posture by platforms stands in direct tension with the open access that startups advocate.

On the hardware side, the show mentioned that Ghost built a computer priced at $3,499, designed to run AI agents around the clock. But the hosts said they do not think it will reach the mass market. This judgment comes from the podcast hosts, not from market data, and whether Ghost's around-the-clock AI agent computer will be accepted by the market remains in question.

In the deals news segment, the panel discussed AI compute company Lambda's reported $4 billion funding round, as well as the $35 billion backlog contributed by a single customer. It should be emphasized that both figures are 'reported' in nature in the coverage and lack first-party source confirmation, so they can only be treated as source claims rather than confirmed facts.

The panel also discussed Uber's $2.3 billion all-cash acquisition of office catering platform ezCater. This deal is not directly related to AI infrastructure, but it appeared in the same episode's topic list.

The podcast discussion also touched on the White House's plan to establish a new 'Super Intelligence Force.' This is another policy-level development, but the source did not explain the plan's specific content or timeline.

Placing this item alongside Amazon and Microsoft abandoning NDAs, one can sense that AI infrastructure is migrating from a purely commercial issue to a public policy issue. But whether there is a direct connection between the two is not supported by evidence in the source, and it should not be overinterpreted.