Mozilla Raises Compliance Concerns Over Google's Prompt API Integration in Chrome

Mozilla has officially registered its opposition to Google's Prompt API, which is currently being tested in Chrome and Microsoft Edge, citing significant compliance and interoperability concerns that could fundamentally alter the open web ecosystem.

The Prompt API Implementation

The Prompt API, as described by Google, "gives web pages the ability to directly prompt a browser-provided language model." This technology enables web applications to send natural language instructions to Google's Gemini Nano model, which runs locally within the browser. The implementation requires substantial storage space—approximately 4.27 GB for the desktop version of the Nano model (v3Nano), though Google recommends having 22 GB available for optimal performance.

Mozilla's Regulatory Position

Jake Archibald, Mozilla's web developer relations lead, articulated the organization's concerns in a GitHub discussion regarding the API. "We continue to oppose this API, and feel it has severe negative consequences to the interoperability, updatability, and neutrality of the web platform," Archibald stated.

Mozilla's opposition centers on three primary regulatory concerns:

1. Interoperability Risks: The fear that Google's Nano model will become the de facto standard, creating pressure for other browser vendors to license it for consistency, thereby undermining the open nature of the web platform.

2. Policy Compliance Issues: The Prompt API requires users to agree to Google's Generative AI Prohibited Uses Policy, which prohibits activities that extend beyond legal requirements, such as generating "disturbing" content. This creates a precedent for vendor-specific rules governing web API usage.

3. Developer Representation Concerns: Mozilla disputes Google's claim of strong developer support for the API, suggesting the company misrepresented the level of enthusiasm through selective evidence.

Compliance Requirements for Web Developers

For organizations developing web applications that might utilize the Prompt API, several compliance considerations emerge:

• Model-Specific Development: Developers would need to tune their applications to the quirks and policies of specific models, potentially creating compatibility issues across different browsers.
• Policy Adherence: Implementing the API requires acceptance of Google's content policies, which may conflict with certain application requirements.
• Performance Expectations: Current testing indicates significant limitations in model performance, with failure rates ranging from 15-30% for various tasks.

Technical Implementation and Alternatives

Web developers currently have multiple pathways to integrate AI capabilities:

• Cloud service APIs for communication with hosted models
• JavaScript runtime frameworks for local model access
• WebGPU and WASM technologies for specialized implementations

Mozilla itself is developing its own AI tools, including a Smart Window feature in Firefox and tools for AI model scaffolding, demonstrating alternative approaches to browser-based AI integration.

Regulatory Timeline and Industry Response

The Prompt API has already progressed to testing in Chrome and Microsoft Edge, despite Mozilla's objections. Google's Rick Byers, the Chrome engineer responsible for shipping the API, acknowledged Mozilla's concerns but emphasized a preference for experimentation and learning from mistakes.

Byers requested evidence of actual harm to advance the discussion, drawing parallels to past debates over controversial web technologies like Encrypted Media Extensions (EME), which ultimately proved less problematic than initially feared.

Performance and Compliance Data

A February 2026 performance report comparing Chrome (Gemini Nano) and Edge (Phi-4 mini-instruct) using the Prompt API revealed significant limitations:

• Generative Tasks: 24.29% of Edge's and 15.17% of Chrome's responses failed to complete assigned tasks
• Classification Tasks: 29.58% of Edge's and 23.93% of Chrome's responses failed to categorize input correctly
• Accuracy Issues: Edge models "hallucinated" 17% of the time, while Chrome models did so 6% of the time

These performance metrics raise questions about the current readiness of the technology for widespread deployment and the compliance implications of relying on such variable outputs for critical applications.

Future Compliance Considerations

As this technology evolves, organizations should consider several compliance factors:

1. Vendor Lock-in Risks: Standardizing on a specific browser AI model may create dependency on a single vendor's technology roadmap and policies.

2. Content Policy Alignment: The requirement to adhere to vendor-specific AI policies may conflict with organizational content guidelines or regulatory requirements.

3. Performance Validation: Implementing AI features through browser APIs requires rigorous testing to ensure reliability and accuracy, particularly for applications where incorrect outputs could have significant consequences.

Google has responded to Mozilla's concerns by stating, "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."

For organizations monitoring this development, the key compliance consideration remains whether browser-specific AI APIs align with their strategic requirements for interoperability, neutrality, and long-term maintainability of their web applications.