News Bites: January 26, 2026

In Neal Mohan’s January 21 letter, YouTube said creators will soon be able to make Shorts using their own AI likenesses, while also expanding AI-assisted creation, auto-dubbing, in-feed image formats, shopping, and brand-deal tools. The company said Shorts are averaging 200 billion daily views, more than 500,000 creators are already using YouTube Shopping, and millions of viewers are now watching dubbed content each day.  

What changed is that YouTube presented AI not as a side toolset, but as part of its main monetization and distribution strategy across creator content, TV viewing, shopping, and brand partnerships. For marketers, that makes YouTube’s AI roadmap a media and commerce story, not just a creator-product update.  

Google said on January 22 that AI Pro and AI Ultra subscribers in the U.S. can opt in to connect Gmail and Google Photos to AI Mode in Search through a new feature called Personal Intelligence. The feature is designed to make responses more tailored for tasks such as shopping and travel, and Google said the experience runs in Labs and gives users control over connected apps.  

This is a meaningful search shift because it pushes Google’s AI answers closer to personal decision support rather than generic retrieval. The commercial significance is obvious: if Search can use personal context to guide recommendations, then the competitive bar for brands rises from keyword relevance to contextual usefulness and product fit.  

Adobe said on January 22 that Firefly Foundry is designed to let studios, brands, and rights holders build proprietary generative AI models trained only on IP they own or control. Adobe said these “omni-models” can support high-resolution images, audio-sensitive video, 3D, and vector output, and positioned the product as a way to speed production while preserving authorship and ownership.  

That is a more commercially useful framing of generative AI for large rights holders than open-ended prompt tools. Adobe is leaning into a market where brand safety, legal clarity, and asset control matter more than novelty, which makes the offering especially relevant to studios, agencies, and global brand teams.  

Anthropic published “Claude’s new constitution” on January 21 and released the full document publicly. The company said the new version is meant to define Claude’s values, behaviour, and reasoning in a more holistic way than the earlier rule-focused approach, including priorities around safety, ethics, compliance, and helpfulness.  

For marketers, this is not just a safety story. It signals that model providers increasingly see behaviour, tone, and normative judgment as part of product differentiation, which matters for enterprise adoption in branded use cases, customer-facing assistants, and regulated communications.  

Members of Congress introduced the bipartisan TRAIN Act on January 22 to create a process for copyright holders to request information about whether their works were used to train AI systems. Supporters said the bill is intended to improve transparency without changing copyright law itself, and both House press releases framed it as a way for creators to seek clarity before costly litigation.  

This is one of the clearest policy signals of the week for creative industries. Even if the bill’s path is uncertain, it reflects a strong push toward training-data accountability that matters directly to publishers, studios, labels, agencies, and any brand concerned about licensed or proprietary content.  

The Verge reported on January 22 that Microsoft is encouraging wider internal use of Anthropic’s Claude Code across major teams, including Windows, Microsoft 365, Outlook, Teams, and others. The report said nontechnical employees such as designers and project managers were also being encouraged to use it for prototyping, while Microsoft continued expanding Anthropic access through Azure and related enterprise tooling.  

The important shift is that enterprise AI tool selection is no longer following a single-vendor path even inside one of OpenAI’s closest partners. That signals a more pragmatic buyer mindset in which enterprises mix models and tools based on workflow fit rather than platform loyalty.  

OpenAI’s January 21 Atlas release notes said the browser added tab groups and a new “Auto” mode that switches between ChatGPT and Google depending on the query. The update also changed how search links are displayed in answers, while outside coverage highlighted the broader positioning of Atlas as an AI-native browser rather than a simple ChatGPT wrapper.  

This matters to marketers because browser behavior is becoming another layer of AI-mediated discovery. If the browser itself decides when to surface web links versus chatbot answers, then traffic patterns, visibility, and the path from question to click may keep fragmenting across interfaces.  

Anthropic said on January 21 that it is partnering with Teach For All to train more than 100,000 teachers and alumni across 63 countries through an AI Literacy & Creator Collective. The company said the network serves more than 1.5 million students and framed educators as co-architects who will both use Claude and help shape how it is applied in classrooms.  

While this is not a direct ad-tech launch, it is a meaningful distribution and adoption story. Large-scale AI literacy programs shape future norms around which tools become trusted, habitual, and institutionally embedded, which matters over time for platform influence and workforce readiness.