CIOs face EU AI Act gaps as enterprises rethink model spending

EU AI Act readiness remains uneven among digital leaders with European interests, even after enforcement was pushed to December 2027. A diginomica network survey found that only 35% of its digital leadership community has started to track, label or watermark AI-generated images, audio, video and text ahead of the August 2026 watermarking deadline. Just 3 percent have completed the work, with 28% still working on it, while 36% had not assessed whether the Act applies to their business.

AI literacy work is further advanced. 38% are part way to compliance with training expectations, 34% are working towards it, and 13% believe they are fully compliant. The regulation is presented as more focused on people than technology, amid broader concerns about sovereignty, external AI providers and dependence on hyperscalers.

Enterprise AI strategy is also shifting away from paying for the latest frontier models by default. Discussion around tokenomics, board priorities and vendor examples points to a preference for the right model for the task, stronger context layers, privileged data and homegrown or localized approaches where they make sense. The caveat is operational complexity: model drift, agentic loop monitoring and context rot require skills and governance many teams may not yet have.