Google Gemini’s rapid AI releases overshadow crucial AI safety assessments and reports.

Google’s Gemini AI assistant is rapidly advancing to keep pace with competitors such as OpenAI’s ChatGPT and Meta AI. The company is introducing new features on a weekly basis, recently launching real-time Gemini Live video capabilities and version 2.5 of the Pro model for free-tier users. However, amidst this swift release schedule, there are growing concerns about the delays in safety reports for its AI models.

As a leader in AI applications on Android, Google faces significant pressure from users and the tech industry to provide AI that effectively aids in streamlining everyday tasks. In an interview with TechCrunch, Tulsee Doshi, head of product for Gemini, acknowledged that the pace of releases is crucial to remaining competitive. Nonetheless, many users may feel uneasy if this speed comes at the expense of safety and security, raising important questions about Google’s priorities.

Reports indicate that Google is behind on safety documentation for models like Gemini 2.0 Flash, which debuted earlier this year. Interestingly, Google initially proposed the idea of evaluating AI models’ safety and other parameters back in 2019. Researchers from the company advocated for “model cards” to accompany AI releases, detailing performance metrics, safety assessments, and potential use cases.

Although these cards are not yet mandated by industry standards, they have been released alongside new AI models in good faith. Google has stated that the latest Gemini 2.5 Pro model does not currently have model cards since it is still categorized as an experimental model. However, there is no such excuse for the older 2.0 Flash model, for which the card documentation is still pending.

The company has previously assured government agencies that safety reports for publicly available models would be provided. Google has confirmed to TechCrunch that the model cards for Gemini 2.5 Pro will be issued once the model reaches public and stable status, as safety and adversarial testing are complete. While users are eager to explore new features, Google’s delay in safety reports may raise concerns among those vigilant about data security.