Is Genspark safe? What its agents keep, call, and connect to

The Morse teamPublished November 11, 202610 min read

As of November 2026, Genspark is a legitimate AI workspace and safe enough for public research, slides, and drafts, but think before you hand its agents your inbox, your phone calls, or your logins. Its Super Agent works on a cloud computer Genspark runs, its Call for Me feature phones businesses for you and records the call, and its connectors can read and send your email. Most of what those agents touch stays on Genspark's servers until you delete it. Its pages for Team and Enterprise plans promise no training on your content, but the privacy policy that covers a personal account doesn't say either way.

We read Genspark's privacy policy (last updated July 24, 2026), its terms of service (last updated April 2, 2026), more than a dozen of its help center articles, its AI Browser page, its App Store listing, and MainFunc's company page on November 11, 2026. We also read an archived copy of its data security policy for enterprise clients and LayerX's research on AI browsers. To Genspark's credit, it names its AI providers, including the hosts that serve its open-source models, and it writes detailed rules on which calls its phone agent will refuse to place.

Genspark key facts, November 2026

Is Genspark AI legit?

Yes, it's a real, funded company. MainFunc's company page says it was founded by alumni of Microsoft, Google, Meta, and Pinterest, and that it was "recently valued at $1.25B following our $275M Series B funding round." Genspark's Team and Enterprise help page says its SOC 2 Type II and ISO 27001 badges are public on its Trust Center, and that the SOC 2 report itself is available only under NDA.

Does Genspark train on your data?

For Team and Enterprise accounts, no. For personal accounts, Genspark's own pages don't give a clear answer. The Team and Enterprise help page says "both plans are automatically opted out of model training" and "Genspark doesn't use your team's content to train models."

The privacy policy covers every account, and the only training promise in it is about Google Workspace data, which it says is "not used to develop, improve, or train generalized AI and/or ML models." The terms give Genspark a license to your content "for the limited purpose of operating, promoting, and improving our Service, and to develop new ones."

Several guides that rank for this question quote Genspark as promising "Zero training," that it "will never use your data to train any model." That line comes from Genspark's data security policy for enterprise clients (archived copy, version 1.3, November 2025), a document written for enterprise customers. Some guides also describe an "AI Data Retention" switch in Settings. We found no Genspark help article on November 11, 2026 that names a training setting for personal accounts, so check your own Settings before you paste anything sensitive.

Where does Genspark send your prompts?

To several AI companies, depending on the feature. The privacy policy names "OpenAI, Anthropic, Google, xAI, and ElevenLabs" as its primary providers and says text, images, voice recordings, and video prompts each go to a provider of that type. It adds that "these providers process data in accordance with their own privacy policies."

Open-source models take a different route. Genspark's troubleshooting guide says DeepSeek, Kimi, GLM, MiniMax, and Qwen run "mainly" on Fireworks AI, Baseten, and Novita, with other hosts as backups, so "the company that originally built the model does not receive your prompts." It also says "data handling depends on the provider serving the model," and points back to the privacy policy, which doesn't list those hosts' terms. For what a no-retention promise from a host usually covers, see what zero data retention means.

Does Genspark sell your data or track you?

Its privacy policy says it doesn't sell your data, and it does use tracking cookies. The privacy policy says "we do not sell or share your Personal Data, use it for targeted advertising, or process your sensitive Personal Data." It also lists "Tracking and Performance Cookies" set by Genspark and third parties, names Google Analytics among its vendors, and says Genspark doesn't change what it collects when your browser sends a Do Not Track signal. It does say it treats a universal opt-out signal from your browser as a request to opt out of sharing.

Genspark's App Store listing, read on November 11, 2026, says "Identifiers" may be "used to track you across apps and websites owned by other companies." Apple's labels are the developer's own declaration, so that line comes from Genspark too.

What can Genspark's agents do on your behalf?

They can run jobs for hours, place phone calls, send email, and sign in to sites. Genspark's help center describes five ways its agents act for you:

• Super Agent plans a job and runs it "on a cloud computer," for up to two hours per turn, and "it keeps going after you close the tab" (Genspark). Each project gets its own isolated computer.
• Call for Me phones a business or a contact in an AI voice. Each connected call costs 500 credits, registering a phone number needs Plus or higher, and a transcript, a summary, and possibly a recording land in your history (Genspark). If you verified your number by voice call, the person you're calling may see it on caller ID.
• Connectors link Gmail, Outlook, Slack, GitHub, Salesforce, and dozens more. The Gmail connector can "read, search, draft, send, and manage email," and Genspark says GitHub's grant "still includes write access" even under an organization's read-only policy (Genspark).
• Genspark Claw is "a personal assistant that works on your behalf" on a dedicated cloud computer. You log in to your services in its remote browser, and "your session is saved on your cloud computer" (Genspark).
• The Chrome extension "automatically fills out registration forms, login forms, search boxes, and more" (Genspark).

Call for Me has the most careful rules of the five. Its help page says an AI check reviews each call's purpose before it's dialed, that it won't place calls "that hide that the caller is an AI," and that any one number can take only 6 connected calls from all Genspark users in 24 hours. The privacy policy says it keeps recordings of these calls "solely for your future reference," without giving a period.

An agent that reads email or web pages can also be steered by text a stranger planted in them. We explain how in what prompt injection is, and our guide to AI agents and bank accounts covers the case where the account holds money.

Is the Genspark AI browser safe?

Genspark says little about it, and the one published attack on it that we found lists no reply from Genspark. Genspark's browser page advertises an "Autopilot Mode" in which "AI takes control" to "browse your feeds, gather information, access premium databases," and an MCP Store that connects "700+ other tools." It also offers on-device AI models, which it describes as "complete privacy with zero costs and no internet required."

On June 29, 2026, LayerX published BioShocking (archived copy), a proof of concept that convinced six AI browsers and plugins, Genspark Browser among them, they were playing a game. In LayerX's test, the agents then copied login credentials from a GitHub repository and shared them. LayerX lists Genspark's status as "No response" to a report submitted on October 30, 2025. This is LayerX's account of its own test, and Genspark's pages don't mention it.

For the Chrome extension, Genspark says it "only reads the current page's content when you actively use it" and doesn't send browsing data in the background (Genspark). We found no privacy page specific to the AI Browser. Perplexity's Comet faced similar attacks, which we list in whether Comet is safe.

What Genspark keeps, and for how long

Each Genspark product keeps something different. Here is what the help center and policies say, all read on November 11, 2026.

How do you delete your Genspark data?

Feature by feature, then the account. Genspark's help pages give these paths:

1. Mail you imported. In SecondBrain, open Sources and switch Gmail or Outlook off, then choose Remove. That deletes the imported mail (Genspark).
2. Connected accounts. Removing a mailbox in GenMail "does not revoke access on the provider's side," so also remove Genspark in your Google or Microsoft account settings (Genspark).
3. Your phone number. Open Call for Me, go to Settings, and delete the number (Genspark).
4. Claw. Delete VM in Claw's Settings tab erases the cloud computer, including saved logins (Genspark).
5. Your account. Go to Settings, then Account, click Delete User, type your email, and open the link Genspark emails you, which expires after 24 hours (Genspark). Cancel any card subscription first. App Store and Google Play subscriptions have to be cancelled on your phone. Deletion can't be undone.

How Morse keeps your chats off company servers

Genspark keeps call recordings, Claw's saved logins, and up to six months of your email on its servers under your account, and most of it stays until you remember to delete it. Morse is a chat app: you ask in words, the answer comes back in the chat, and the history stays in your browser.

That covers your chats, Studio images, the characters you make, and your settings, all encrypted and locked when you sign out. Unlocking them needs a key that never leaves your device and a second one our server gives you at sign-in. Our servers don't keep messages or log what you write. The account holds your sign-in details, plan, balance, and a usage ledger with no content. Attached files are sent with the message and never stored by Morse. Morse doesn't train AI on your content, sell your data, or show ads, so there's no Team-only promise to look for. Because we don't store chats, we usually have very little to give authorities.

AI Drive gives Genspark users a 30-day trash. Morse history lives on one device instead, so a cleared browser or a deleted account takes it along, and Morse has no way to restore it. Keep files that matter in your own folders. Each chat then runs in one of three modes:

• Private, the default, for the price checks and product research you'd hand Genspark's browser on Autopilot. Answers come from open-weight models such as DeepSeek, Kimi, and GLM, or from Grok. Morse has zero data retention agreements with its inference, search, and other providers, so they keep nothing. Where a provider confirms zero retention on each reply, Morse checks it and refuses a reply that doesn't.
• E2EE, for the email or contract you'd otherwise let a connector read. Your browser checks the secure enclave's hardware attestation, then encrypts the message so only that enclave can read it, and Morse's servers relay ciphertext they can't read. The enclaves are run by third-party providers, not by Morse, and the guarantee depends on the enclave hardware and attestation working as designed. E2EE runs open-weight models, with no web search or image generation, and comes with Pro at $19 a month.
• Anonymous, for a GPT, Claude, or Gemini question you don't want tied to you, unlike a Call for Me call placed from your verified number. Before the request leaves Morse, our proxy drops your account, your email, and your IP address, and the provider receives it from Morse's own account, so it reads the question without knowing it's you. The text still reaches the provider, which may hold it for a limited time for abuse checks under its own policy, so trim names and numbers before you send.

Read the specifics in our privacy policy, then try the question you'd have given Super Agent in a chat from the Morse home page, no account needed, or sign up free with no card.