In today’s Digest, we look at the Trump administration considering AI model safety reviews, Duncan Painter leaving Omnicom’s data and commerce role, and Image AI models outpacing chatbots in app growth

US rethinks AI safety stance

The Trump administration is weighing a new AI security framework that would require the Pentagon to lead safety testing of AI models before they are deployed across federal, state, and local government systems. The proposal, discussed by the White House’s Office of the National Cyber Director in recent meetings with tech firms and industry groups, reflects growing concern over the security risks posed by advanced AI systems. 

Officials are also considering whether to formalise an executive order that could assign multiple agencies responsibility for evaluating AI systems before public-sector deployment. Sources say the framework is already at an advanced stage of development, though White House officials stress that no final policy has been approved. 

The renewed focus on AI safety comes amid heightened cybersecurity concerns within the national security community, while economic and technology stakeholders warn that stricter testing requirements could slow down deployment and innovation. 

The plans contrast Trump’s previous demands surrounding AI safety, including signing an executive order which sought to prevent states from enforcing their own AI regulations.  

Omnicom data chief steps down

Duncan Painter is stepping down from his role at Omnicom Group, where he led the Omni platform and Omnicom Commerce, marking a leadership change as the company integrates its recent acquisition of Interpublic Group. Painter, who became a central figure after Omnicom’s USD$835m (£659.7m) acquisition of Flywheel in 2023, is leaving to pursue a senior executive role in the UK, citing a desire to be closer to his family. 

In response, Omnicom has reshuffled its leadership structure, appointing Christine Gambino as CEO of the Omni platform, while Alex McCord will assume expanded oversight of Omnicom Commerce alongside his current responsibilities. 

Image AI models outpace chatbots in app growth

A new report from app intelligence firm Appfigures shows that image model releases are now a major driver of growth for AI mobile apps, generating about 6.5 times more downloads than traditional model updates. The data suggests a shift in user behaviour, with visual-generation features overtaking conversational upgrades as the primary trigger for new installs, even as earlier growth phases were fuelled by advances in chat models and voice interfaces.

The trend is evident across major platforms. ChatGPT saw more than 12 million additional installs in the 28 days following the rollout of its GPT-4o image model, significantly outperforming previous model releases. Similarly, Google Gemini recorded over 22 million incremental downloads after introducing its Nano Banana-powered image capabilities within the Gemini 2.5 Flash update, more than quadrupling its install rate during that period. 

Smaller but notable gains were also observed with Meta AI, where the launch of its Vibes AI video feed added an estimated 2.6 million downloads, underscoring the growing pull of visual-first AI features in mobile adoption.