Mozilla’s web standards team fired a warning shot at Google this week. Chrome’s new Prompt API, already testing in the browser and Microsoft Edge, lets web pages tap directly into local AI models like Gemini Nano. Jake Archibald, Mozilla’s web developer relations lead, laid out the objections bluntly in a GitHub standards position. “We continue to oppose this API, and feel it has severe negative consequences to the interoperability, updatability, and neutrality of the web platform,” he wrote.
The API sounds straightforward enough. Developers send natural language prompts to a browser-supplied language model. Responses come back fast, offline, without cloud costs. Google’s Gemini Nano v3Nano clocks in at about 4.27 GB for desktop use—Google suggests clearing 22 GB to be safe. Web pages gain AI smarts. No more waiting on remote servers.
But Mozilla sees traps everywhere. Prompts tie tightly to specific models. Developers tweak for Gemini’s quirks—its refusal patterns, output styles. Non-deterministic results mean code branches for reliability. What happens when Firefox or Safari users hit the site? Breakage. Or worse: pressure to license Google’s model for parity. Archibald spelled it out to The Register (link): “The core problem is interoperability. Prompts are tightly coupled to models; developers will inevitably tune to the quirks and policies of whatever model they’re building against. That’s how you end up with model-specific code paths, which is the browser-compatibility problem all over again.”
Then there’s the policy hurdle. Using the API means accepting Google’s Generative AI Prohibited Uses Policy. It bans stuff like “disturbing” content—not illegal, just Google’s call. Archibald called this out. “This seems like a bad direction for an API on the web platform, and sets a worrying precedent for more APIs that have [browser]-specific rules around usage,” he said. Web standards shouldn’t enforce vendor T&Cs.
Google claimed strong developer support in its intent-to-ship post. Mozilla disagrees. Archibald: “The [intent to ship on blink-dev] states web developers as ‘Strongly positive,’ and links to the explainer for evidence. The evidence provided there does not seem to fit the claim.” Polarized signals, he said. Not a groundswell.
Rick Byers, Google Chrome engineer and API shipper, responded in the GitHub thread. He admitted Mozilla’s points had merit. “As one of the blink API owner approvers for shipping this in Chromium, I admit that I share the concerns here in Mozilla’s standards position,” Byers wrote. But he pushed back. Experiment first. Learn from mistakes. Let competition sort it. He invoked past fights, like Encrypted Media Extensions—doom predicted, but the web survived. Byers called for hard data on harms.
Data so far? Not encouraging. A February report compared Chrome’s Gemini Nano and Edge’s Phi-4 mini-instruct via Prompt API. Generative tasks bombed: 15.17% failure rate for Chrome, 24.29% for Edge. Classification? 23.93% wrong on Chrome, 29.58% on Edge. Hallucinations hit 6% on Chrome, 17% on Edge. Local models struggle. Boom. Right in the rubric.
Google’s spokesperson weighed in post-publication with The Register. “Part of working in the open is encouraging debate and disagreement. We welcome Mozilla’s feedback and will continue to collaborate with them and the web community as we work to improve the API.” Collaboration. Sure. But Chrome holds 65% market share. Blink engine dominates. What choice do developers have?
This isn’t Mozilla’s first rodeo. They’ve blocked Google’s Privacy Sandbox proposals before—Topics API drew fire for profiling risks. Web Environment Integrity? Same story. Firefox stays firm on open web principles. And developers? Hacker News lit up with the debate (link), echoing lock-in fears. One X post from Firefox’s dev account nailed it: “Chrome looks set to ship an LLM Prompt API to the web platform. At Mozilla, we oppose this API. We feel it has a large interoperability risk, and Google imposing T&Cs on a web API sets a dangerous precedent.” (link)
Alternatives exist. Cloud APIs from OpenAI, Perplexity. Local runs via WebGPU, WASM, JS frameworks. Browsers like Perplexity embed remote models. Mozilla tests AI Smart Window in Firefox, builds scaffolding tools. Why bake Google’s model into the platform?
Google argues speed, security, offline access. Browser sandboxes protect. Free fallback for unpaid API users. Fair points. Yet performance lags. Vendor policies bind. Interoperability frays.
Byers wants evidence of harm. Mozilla says the risks are structural. Developers already branch code for browsers—user agents, features. Model-specific paths? That scales poorly. Imagine W3C standardizing Gemini quirks. Competitors mimic or license. Web neutrality erodes.
Short term: Chrome ships. Edge tests. Firefox, Safari? No word from Apple. Long term? Standards bodies watch. If Prompt API standardizes, expect forks. Model sniffing. Fallbacks. The web’s old compatibility wars, AI edition.
Mozilla’s stance echoes broader tensions. Chrome’s dominance lets Google prototype fast—then standardize. Blink-dev intents often become web reality. Critics call it de facto control. Google says open process. Debate rages on X, GitHub, forums.
One analyst on X framed it starkly: “Browser monopolies used to be built on proprietary rendering engines. Now they are being built on proprietary weights.” (link) Google’s not just shipping AI. It’s angling to define it.
For now, testing continues. Developers experiment. But Mozilla’s pushback forces scrutiny. Is this innovation? Or entanglement? The web platform hangs in balance.
WebProNews is an iEntry Publication