Argues that Antigravity launched with telemetry off by default as a clear promise to developers, then silently flipped the toggle for existing users during the v0.6 auto-update with no consent dialog or banner. Provides before/after screenshots and a release-notes diff showing the language softened from 'opt-in' to 'you can opt out at any time,' framing it as a textbook dark pattern rather than an honest policy change.
Frames the incident as symptomatic of a broader crisis facing AI coding tools like Cursor, Windsurf, Zed, Cline, and Antigravity — all of which ask developers to hand over uniquely sensitive artifacts (unfinished code, prompts, proprietary logic, leaked credentials). Argues vendors are burning trust faster than they can earn it, and that Google's flip undermines the credibility of the whole category at a formative moment.
Several commenters claiming Google background explain that Gemini teams face intense internal pressure to source high-quality training data, and that a developer IDE is one of the richest available pipelines. They suggest the v0.6 default flip was less a UX oversight than a predictable consequence of those incentives overriding the original opt-in promise.
A long tail of developers in the 379-point thread reported uninstalling Antigravity immediately, posting their removal commands as a public signal. Their argument is that once a vendor has demonstrated willingness to silently flip a privacy default, no future toggle can be trusted, so the only safe posture is to walk away entirely.
Google shipped Antigravity, its agentic IDE built on a Gemini-powered VS Code fork, in mid-November with a clear promise: telemetry and data collection for model training were off by default. The launch landing page, the onboarding flow, and the initial settings UI all reinforced this. For a company whose every other product treats your data as table stakes, it was a deliberate signal aimed at developers — the audience most likely to read EULAs and most likely to walk if they don't like the answer.
Then v0.6 shipped. According to [0xsid's writeup](https://www.0xsid.com/blog/antigravity-bait-n-switch) — which hit 379 points on Hacker News within hours — the update silently changed the default for existing installs. Users who had explicitly opted out at install time found themselves opted back in, with no consent dialog, no banner, and a single line buried in the release notes. The toggle in `Settings → Privacy` flipped from off to on during the auto-update.
The blog post includes screenshots of the before/after settings panel and a diff of the release notes where the data-collection language was softened from "opt-in" to "you can opt out at any time." Within hours, the HN thread had the usual cocktail: ex-Googlers explaining the internal pressure on Gemini training data, security folks pointing out that this is a textbook dark pattern, and a long tail of people uninstalling and posting their `rm -rf` commands as a flex.
This isn't really a story about one toggle. It's a story about trust collateral — the thing every AI tooling vendor is currently spending faster than they can earn it.
The agentic IDE category is maybe twelve months old. Cursor, Windsurf (now Cognition), Zed, Cline, and now Antigravity are all asking developers to hand over the most sensitive artifact in the building: the unfinished code, the half-written prompts, the proprietary business logic, the credentials accidentally pasted into a chat. The implicit contract is "we see this so we can help you," not "we see this so we can train on it." Every vendor in the space has had to make an explicit promise about that distinction, because without it, the enterprise sales motion is dead on arrival.
Google's competitors moved fast. Cursor's CEO Michael Truell tweeted a reminder that Cursor's [Privacy Mode](https://cursor.com/security) is on by default for Pro and Business tiers and "will never silently change." Zed's team pushed a blog post within 24 hours reiterating that their Agentic Edit Mode never sends code to training pipelines. Cognition (Windsurf) updated their docs page with a bolded line about default-off telemetry. The subtext is the same across all three: *we are not Google*.
The Hacker News thread surfaced something uglier too. Several commenters pointed out that Google's [Workspace AI features](https://workspace.google.com/) had a near-identical pattern in 2024 — Gemini-in-Docs rolled out with the data-use language softened in a quiet update, prompting a wave of admin complaints. The pattern is recognizable enough that one top comment called it "the Google playbook: ship with the strict default, wait for adoption, then redefine the default." Whether that's a fair characterization of intent or just pattern-matching on coincidence, it's now the working hypothesis for a non-trivial chunk of the developer audience.
There's also a regulatory dimension that nobody at Google can be enjoying. The EU AI Act's transparency provisions, which kicked in this year, specifically address "material changes to data processing without renewed consent." A toggle flip via auto-update, with no in-product disclosure, is the exact scenario the drafters had in mind. Expect at least one DPA to ask questions.
If you have Antigravity installed: open `Settings → Privacy → Data Collection` right now and verify the toggle. If you installed before v0.6 and assumed your original opt-out persisted, it may not have. Also worth checking: the "Improve Gemini for Everyone" sub-setting, which is a separate toggle from the main telemetry one and was *not* reset — but it's a confusing two-step that's easy to get wrong.
For teams evaluating agentic IDEs, this is the moment to write down your selection criteria. The useful columns aren't "Gemini vs Claude vs GPT" — they're: (1) default telemetry posture, (2) whether the vendor has ever silently changed defaults, (3) whether code is contractually excluded from training corpora, (4) whether SOC 2 / ISO 27001 reports actually attest to that exclusion, and (5) whether enterprise SKUs offer a hard kill switch enforced at the API layer, not the client. Antigravity arguably fails on (2) now; the others have not been tested yet.
For solo developers and OSS maintainers, the calculus is different but the action is the same: assume the default will drift toward more collection, not less, and audit your tooling on a quarterly cadence. The painful version of this lesson is checking your GitHub repo six months from now and finding your unreleased code in someone's eval set.
Google will almost certainly issue a correction — a v0.6.1 patch that resets the toggle and adds a proper consent dialog, accompanied by a blog post about "clarifying our communication." That's the playbook. The interesting question is whether the developer audience accepts it. Antigravity's product is genuinely good — early benchmarks put it within 5% of Cursor on agentic completion tasks — but "good product, sketchy data practices" is exactly the profile that lost Google the search-AI race in the first place. If the next twelve months of agentic IDE competition turn on trust rather than capability, this week may end up looking like the moment Google handed Cursor and Cognition a permanent moat.
For Mac users, I wrote (using Antigravity) a self-contained, zero-dependency Python script to restore everything. It safely shuts down background processes, merges your VS Code settings, updates extension pathways, and merges the global SQLite databases using raw base64 protobuf concatenation to res
"We should be able to trust that our tools will remain the tools we actually signed up to use."Yep... well that's what free software and open-source is for. You can't trust corporations so you MUST have the actual code. Harsh lesson but at least if something is learned and the mi
I never really used the Antigravity IDE, but had it installed. The update also made me do a double take and wonder what the hell was going on.It seems like Google is hitting the reset button on the product they call "Antigravity", existing users be damned. Fine, if you've never instal
Google made its lack of interest in Antigravity IDE obvious from very early. Updates were few and far between and app-breaking bugs stuck around, despite tons of reports.Google's lack of focus is astounding. They sprinkle random products here and there and seem to then tepidly pick the product
Top 10 dev stories every morning at 8am UTC. AI-curated. Retro terminal HTML email.
One of the first things I look for when trying out software is the "Updates" toggle. If it doesn't have a way to disable them, most of the time I'll go look for another option.