EU Labels, New Models, Big Shifts
3 Minute Read
Weekly AI News
Labeling rules for AI content take effect August 2, and the standards for how that labeling should look are now settled. At the same time, the industry is shifting from single AI tools toward agents that handle whole workflows, and the growing choice of models keeps lowering the cost of producing images and video.
New EU Rules on AI Labels
On July 8, the European Commission confirmed that a voluntary code of conduct covers the transparency duties in Article 50 of the AI Act. The European AI Board followed a day later. Starting August 2, deepfakes and AI generated or AI edited text on topics of public interest will need clear labels, and the new code sets out practical standards for how those labels should look.
The EU's Cloud and AI Development Act is also set to publish this week, adding more weight to the same push toward transparency around AI generated content.
Networks Rebuild Around Agentic AI
Serviceplan is bundling its enterprise offer under "Plan.Net Agentic AI" and bringing in an AI lead from Deloitte to run it.
Dentsu Germany just created a new Chief Transformation, Data and Delivery Officer role, after folding its digital skills into two new units the week before.
Segmenta Experience is strengthening its leadership team for the same reason.
The pattern is the same across all three: large organizations are organizing around AI agents that run whole workflows, not around single tools.
Three Models Worth Knowing About
META Muse Image
Meta released Muse Image, its first in house image model, free inside the Meta AI app, WhatsApp, and Instagram Stories.
It handles complex prompts, merges several photos into one result, and takes markup edits like circles or sketches directly on the generated image.
Meta plans to bring the same technology into Advantage Plus, its automated ad creation tool.
OPENAI GPT-5.6
OpenAI introduced its GPT-5.6 family in limited preview: Sol for complex work like coding, Terra for high volume tasks, and Luna for quick, cheap drafts and summaries.
Sol scored 80 on the Artificial Analysis coding index, a new high, while using roughly a third less compute than comparable models.
OpenAI also launched ChatGPT Work, an agent built to carry out full tasks rather than single answers.
Nano Banana 2 lite & Gemini omni flash
Google opened up two models for developers: Nano Banana 2 Lite, a fast, low cost image model, and Gemini Omni Flash for video generation and conversational editing.
Both embed SynthID watermarks, which ties directly back to the EU labeling rules above. Video length is still capped at 10 seconds for now.
More Models, More Options
Creative Cloud
Adobe added Kling 3.0 and Kling 3.0 Omni to Firefly's partner model lineup, adding smart storyboarding and multi shot camera control.
Firefly now carries models from Runway, FLUX, Luma AI, Google, OpenAI, Ideogram, Pika, Topaz Labs, ElevenLabs, and Moonvalley under one subscription.
Mistral AI
Mistral AI confirmed early access to a new open weight model family this month, backed by a four billion euro investment in data centers in France and Sweden.
The company is positioning itself as a European alternative to US providers, at a time when data sovereignty is coming up more often in conversations about AI.