Two developments this week that matter for Dutch professionals using AI: on 13 August Google launched Gemini 3.7 Flash, a fast and relatively inexpensive model for coding and agentic workflows, and from 2 August the first AI Act transparency obligations take effect. Both are of practical significance if you are now making decisions about model selection and compliance.
Gemini 3.7 Flash: faster, cheaper, better at code
Google launched Gemini 3.7 Flash three weeks after its predecessor, Gemini 3.6 Flash. The model has not been retrained but improved through algorithm adjustments and user feedback, which explains the rapid follow-up. The price is $0.75 per million input tokens and $3.75 per million output tokens, applying as an introductory rate until the end of 2026. That is half of what Gemini 3.6 Flash cost at launch, according to 9to5Google.
On benchmarks the improvements are concrete: the software development score on DeepSWE v1.1 rose from 49.0% to 65.3%, and AutomationBench (business workflows) went from 17.0% to 30.4%. For legal teams and financial service providers processing documents: the GDP.pdf score rose from 22.0% to 34.0%.
The model is available via the Gemini API, Google AI Studio, Android Studio and the Gemini Enterprise Agent Platform. For individual users it is available in the Gemini app via the Spark feature, which requires an AI Pro or Ultra subscription. The context length remains 1,048,576 tokens, identical to its predecessor.
By comparison, Claude Sonnet 5 costs $2/$10 per million tokens and scores higher on reasoning, while Gemini 3.7 Flash now comes closer on coding-specific benchmarks. For development teams running large volumes of code generation, the lower price is relevant at comparable quality for that specific task. See the model overview for a direct comparison.
AI Act transparency obligation: chatbots and synthetic content
From 2 August 2026 the first concrete obligations under the AI Regulation (article 50) apply. NLDigital explains the five categories: organisations deploying chatbots must inform users _before the interaction_ that they are communicating with AI. Synthetic content, text, image, audio, video, must be marked in a machine-readable way as AI-generated.
Two nuances that are practically relevant:
Transition period for existing systems. Generative AI systems already on the market before 2 August have until 2 December 2026 to become compliant. If you are now using an AI system that is already running, you therefore still have some time, but not unlimited time.
Who is responsible? For communicative AI (chatbots), the provider is responsible for the transparency notice. For deepfake technology and emotion recognition, the deployer (your organisation) is responsible for informing those affected. That distinction is practically relevant if you use models via API from major providers.
The Dutch Data Protection Authority (Autoriteit Persoonsgegevens) advises signing the European code of practice and using the published EU icons for labelling. This is not a GDPR obligation but an EU AI Act obligation, both can apply at the same time if you process personal data.
For sectors such as healthcare and HR, where emotion recognition or biometric categorisation is conceivable: check whether your systems fall into those five categories. Enforcement by national supervisory authorities, including the AP, is active.
What does this mean in combination?
Gemini 3.7 Flash offers a cheaper option for code generation and document processing. Google processes data in the US; for Dutch organisations in regulated sectors, the data processing agreement framework applies here too. On the safety page you can see which models offer a GDPR-compliant data processing agreement.
The transparency obligation from 2 August makes setting up AI use more formal: you can no longer simply deploy a chatbot or content generator without informing users. For organisations only now starting with AI implementation, this is a good moment to set it up correctly from the outset rather than correct it afterwards.
Sources: 9to5Google, Gemini 3.7 Flash launch (13 Aug 2026) · NLDigital, AI Act transparency obligations · Autoriteit Persoonsgegevens, AI transparency from 2 August · Miraflow, Gemini 3.7 Flash benchmarks