AI tools that shut down or merged in 2026: what happened and what to do

AI tools that shut down or merged in 2026: what happened and what to do cover graphic

AI products can disappear quickly. In 2026 we saw a workspace startup close after its team joined a large company, a podcast app shut down a day after a big platform launched a similar feature, a browser product folded back into the assistant it came from, and an agent product wind down within weeks of an acquisition announcement.

This article lists the cases we could find in dated sources, notes what each report says about customers and their data, and offers a plan for reducing the damage if a tool you rely on closes. It is a news summary and a practical guide. It is not a list of every closure, and we could not verify every claim on a vendor’s own page.

Key takeaways

  • Several 2026 closures followed acquisitions where the buyer wanted the team, not the product, such as Cove (closed 1 April) and Vercept’s Vy (closing 25 March), according to press reports.
  • Some shutdowns gave users only weeks. Agency’s product was reported to end on 31 August, 26 days after the acquisition was announced.
  • One publisher that tracks acquisitions reports that about 36 percent of the 229 acquired AI tools it follows were folded into the buyer, and warns its figures cover only its own catalogue.
  • Reports on several shutdowns say details about data export and deletion were thin or not published.
  • Keep exports, avoid deep lock-in and read the terms on termination before you build a workflow around any AI tool.

Summary

The reports we read describe three recurring stories. In the first, a small startup is acquired and its product is closed within weeks while the team moves to the buyer. In the second, a large platform adds a similar feature and a standalone product loses its reason to exist. In the third, a company simply fails to find enough users or cannot cover its costs. The common risk for a customer is the same: your workflow depends on something with a short notice period and an unclear data plan.

We use the words “reported” and “according to” throughout because several of these items come from news summaries and aggregators, and we could not open every original page. Where a source gives no reason, we say so.

Timeline of reported shutdowns and mergers

Date (2026) Product or company What was reported Source type
26 February (report date) Vercept (Vy), acquired by Anthropic Vy to close on 25 March as the team joins Anthropic; deal terms not disclosed News report (Techeconomy)
18 March (report date) Cove, acquired by Microsoft Software to shut down on 1 April; team of fewer than 10 joins Microsoft Directory summary of a TechCrunch report
March Sora (OpenAI video app) Shut down, reportedly due to high operating costs and retention challenges AI Weekly summary of a TechCrunch list
22 May (report date) Huxe AI podcast app shutting down; removed from app stores and data deleted after a grace period Aggregator summary of a TechCrunch report
5 August Agency (Kai), acquired by Klaviyo Acquisition announced; product to wind down on 31 August Analysis article (Beri)
August Relay (workflow automation startup) Closed, reportedly because large platforms built similar automation AI Weekly summary
9 August ChatGPT Atlas (OpenAI browser) Discontinued, with useful features absorbed into ChatGPT AI Weekly summary

Acquisitions where the product closed

Vercept and Vy

Techeconomy reported on 26 February that Anthropic acquired Seattle-based Vercept, whose tools help AI systems carry out multi-step tasks on a computer. Financial terms were not disclosed. Its main product, Vy, which let a cloud-based agent operate a remote Apple MacBook, was to close on 25 March, and the report’s headline says customers had 30 days to exit. Some Vercept founders were reported to be joining Anthropic while others were staying behind.

Cove and Microsoft

A directory summary of a TechCrunch report dated 18 March says Cove, a Sequoia-backed startup that turned AI chats into a visual workspace, would shut its software on 1 April while its team of fewer than 10 joined Microsoft. It had raised $6 million in seed funding in 2024. The summary says there was no public, detailed plan for customer data and that users should contact Cove directly about export and deletion.

Agency and Klaviyo

Beri’s analysis reports that on 5 August Klaviyo announced it was acquiring Agency’s team and technology, with the deal expected to close in the third quarter. Agency’s homepage is quoted as saying the company would wind down on 31 August. The article notes that the press release and trade coverage did not mention a shutdown, and that no export procedure, retention commitment or refund policy had been published. It observes that Klaviyo sells to consumer brands while Agency’s Kai served B2B software companies, so there was no equivalent product to migrate to. The article’s advice to customers is the author’s opinion, not a documented fact.

Products absorbed by big platforms

AI Weekly, summarising a TechCrunch running list of AI projects that did not last, reported on 25 October that ChatGPT Atlas, OpenAI’s standalone web browser, was discontinued on 9 August after less than a year, with its most useful features absorbed into ChatGPT. It also lists Sora as shut down in March 2026 and says it struggled with high operating costs and user-retention challenges. We could not open the TechCrunch list itself, so we treat these as second-hand reports.

The pattern behind Atlas is worth noting for buyers. A product from a large vendor is not automatically permanent. When features move into a flagship product, the standalone version can go.

Standalone startups that closed

Huxe

An aggregator summary of a TechCrunch article dated 22 May says Huxe, an app that turned a text prompt into a podcast, was ceasing operations. It would be removed from major app stores and user data deleted after a brief grace period. The announcement came one day after Spotify rolled out a similar AI personal-podcast feature. Huxe never disclosed a specific reason, and the summary suggests market pressure from other companies adding text-to-audio features. That last point is the summary’s inference.

Relay

AI Weekly says Relay, a five-year-old workflow automation startup, closed in August 2026 and that as OpenAI, Google and others built similar automation into their own tools, it struggled to justify itself as a standalone product.

How common is a shutdown after a deal?

ToolDirectory.ai, a directory publisher, reported as of 10 November that of 229 AI tool acquisitions it tracks, 147 (64.2 percent) still ship under their own brand and 82 (35.8 percent) were folded into the acquirer and retired. A rebrand into the parent counts as retired in its method. Its figures by acquisition year are below.

Acquisition year Still shipping under own brand Count
2026 76.5 percent 62 of 81
2025 60.8 percent 48 of 79
2024 51.2 percent 21 of 41
2023 57.1 percent 8 of 14

The publisher cautions that recent years look better partly because those tools have had less time to be shut down, and that the figures cover only acquisitions in its own catalogue, not the whole market. It also lists a 2022-and-earlier group with the same numbers as 2023 and flags that this should be verified. Treat the table as a rough illustration, not a forecast. The same source says it tracks hard shutdowns separately and excludes them from this analysis.

What it means for tool buyers

  • Small teams carry more closure risk. Several of the cases involve startups with small teams, though a large vendor’s standalone product can also be retired, as Atlas shows.
  • Notice periods can be short. The reports mention 30 days for Vy and about 26 days for Agency’s Kai.
  • Data plans are often unclear. Several summaries say export and deletion details were missing or not published.
  • Acquisition news is a warning sign, not a guarantee. The tracker data suggests many acquired tools keep shipping, but it also shows a substantial share do not.
  • Free and cheap tools offer less leverage. If you have no contract, you have few rights to notice or refunds beyond what the terms say.

For vendor checks before you buy, see how to evaluate AI tool vendors on privacy, which covers deletion and retention questions.

An exit plan for AI tools

You can reduce the risk before anything happens. The table gives a simple plan, and it works best if you set it up during a pilot, as described in how to run an AI tool pilot in 30 days.

When Action Why
Before you buy Check whether you can export your content and settings, and in what format An export you cannot open later is little help
Before you buy Read the terms on termination, notice and refunds These decide what you can claim if the product closes
While you use it Keep your prompts, templates and key outputs in your own files Your know-how should not live only in the tool
While you use it Name an alternative and note what a switch would involve A fallback shortens the scramble
On an acquisition announcement Export everything immediately and ask the vendor in writing about data deletion Notice periods can be weeks
On a shutdown notice Confirm deletion dates, refund eligibility for unused prepaid fees and migration help Get answers in writing

Contracts, data and refunds

If the tool holds company or customer information, a shutdown is also a data-handling event. Ask the vendor how and when your data will be deleted, and keep a record of the answer. Our guide to protecting company data when using AI tools covers what to avoid putting into tools in the first place.

On refunds and contract rights, rules vary by country and by contract, and what you can claim depends on your terms. This is general information, not legal advice. If a significant sum or sensitive data is involved, speak to a qualified adviser.

What to watch next

  • Whether more standalone agent and workflow products are absorbed as large platforms add similar features.
  • Whether acquirers publish clearer migration and data-deletion plans when they close a product.
  • How the ToolDirectory figures change as the 2026 cohort ages, given the publisher’s own caution about recent years.

These are questions to watch, not predictions.

Frequently asked questions

Does an acquisition mean a tool will shut down?

No. One tracker reports that most acquired tools it follows still ship under their own brand, though a sizeable share do not. Treat news of a deal as a prompt to export your data and check the vendor’s plans.

How much notice will I get?

It varies. The cases we found mention about 30 days for Vy and about 26 days for Agency’s Kai. Your contract or terms may say more.

What should I export?

Your content, templates, prompts, settings, integration details and any logs you may need for records. Ask the vendor what formats are available.

Can I get a refund for unused time?

It depends on your terms and local rules. Ask in writing. This is general information, not legal advice.

Are large vendors’ products safer?

Not automatically. The reports include products from large vendors that were discontinued or absorbed, such as Atlas and Sora, although the original pages were not available to us.

Conclusion

The 2026 closures we found share a pattern: short notice, unclear data plans and a product that was easier to fold into a bigger platform than to keep separate. You cannot predict which tool will go, but you can limit the cost by keeping your own copies, reading termination terms and having a fallback. For a wider selection process, see how to choose an AI tool.

Sources


Keep reading