Google has patched several bugs in its Gemini app that caused users to rapidly exhaust their usage quotas. The fixes include doubling video generation limits for Ultra members and eliminating charges for failed requests.
Google addressed critical billing issues affecting Gemini users this week. The primary bug caused single or double video generations to consume an entire monthly usage quota, frustrating subscribers who expected significantly more access.
The company has implemented three key fixes:
Quota adjustments: Ultra members now receive double the monthly video generation allowance, effectively providing more value for their subscription tier.
Failed request billing: Google will no longer charge users for requests that fail to complete. This prevents users from paying for unusable content and removes a major source of quota waste.
Transparency improvements: The company plans to enhance how it communicates quota usage and limits, giving users clearer visibility into their remaining access and how different features impact their allocation.
The bugs appear to have stemmed from how the system counted usage against quotas. Video generation, which requires significant computational resources, was being disproportionately charged against user limits. The fixes suggest Google's pricing model may not have accurately reflected the actual resource consumption of certain features.
These changes arrive as Google continues expanding its Gemini suite, which now includes Spark, a new AI agent designed to handle multi-step background tasks. The quota fixes are intended to make Gemini's paid tiers more attractive while addressing legitimate user frustrations about value.
Google has not disclosed how widespread the quota bug was or how many users were affected. The company also did not announce compensation for users who may have burned through quotas prematurely before the fix was deployed.
Daydream CEO Julie Bornstein says AI-powered shopping is becoming a critical test case during back-to-school season. The startup's platform lets users describe themselves and chat with an AI agent to receive personalized product recommendations.
GPT-6 Astra achieved 62.7% on the ARC-AGI-3 benchmark using standard testing parameters, significantly outperforming competitors. With a specialized provider adapter harness, the model reached 99.9% accuracy.
Anthropic has confirmed that Claude is experiencing an outage affecting multiple AI models. Users are reporting elevated error rates when attempting to send requests to the service.
OpenAI CEO Sam Altman announced Astra during Bloomberg Television's market close coverage on September 3, 2026, describing the system as a significant advancement in the company's progression toward artificial general intelligence.