UK Games Expo bans generative AI art, text and merchandise across NEC halls

A major UK convention closes the door on synthetic artwork

UK Games Expo has banned generative AI art, text, and merchandise across all NEC exhibition halls, bringing the convention in line with broader British rules on synthetic content. The move affects exhibitors who had relied on AI-generated booth art, marketing copy and printed merchandise; they now need human-authored substitutes or face rejection from the show floor. For generator-side practitioners, the ruling sharpens the gap between consumer-facing and prosumer-facing venues, where provenance and AI-labelling rules are tightening fastest.

Vendors shipping AI art tools into events, tabletop publishers and convention retailers should expect similar conditions to appear in future exhibitor manuals.

Adobe pushes agentic features and a new Firefly model into creator apps

Adobe launched new agentic AI features in Photoshop and Express, including the Firefly Image 5 model and a soundtrack generator for creators. For practitioners, the practical change is that Photoshop and Express now expose automated, multi-step editing flows rather than single-prompt generations, and the Firefly Image 5 model sets a new floor for in-app image quality. The bundled soundtrack generator also moves audio production into the same surface as image work, which matters for users who assemble creator assets in one pipeline.

Teams evaluating Firefly Image 5 should benchmark it against their current image model on the asset classes they ship most, since licensing terms and watermark behaviour typically shift with each Firefly revision.

OpenAI Academy reaches a two-year mark with a broader community focus

Marking two years of OpenAI Academy and bringing AI skills to even more communities, OpenAI is signalling an expansion of its free training programme rather than a single product drop. The implication for practitioners is that onboarding pathways for ChatGPT, the API and adjacent tools are being widened, which typically produces more qualified end users and a steadier stream of community-built prompts and fine-tunes. Teams that publish developer documentation or run internal training should expect an increase in applicants who have already completed Academy modules and reference that credential directly.

Google Beam widens its footprint with new regions and partners

Google Beam expands with new regions, partners, and customers, extending the reach of its communication platform beyond the original launch footprint. For practitioners shipping multilingual or remote-first products, a wider Beam rollout changes the default video-meeting surface that enterprise customers request integrations with. The expansion also signals that Google is treating Beam as an infrastructure play rather than a consumer app, which means SDK and API stability are likely to receive more investment in the coming release cycle.

DeepMind adds server-side memory to Private AI Compute

Introducing private, server-side memory to Private AI Compute for personal AI, DeepMind is moving user state off-device while keeping it inside a sealed compute environment. The change matters to anyone building personal assistants on Google's stack, because persistent context has historically required either a client-side store or a third-party database; server-side memory collapses that into the platform. Privacy-sensitive deployments in healthcare, legal and education, where on-device memory was a deal-breaker, now have a cleaner path to long-running assistants.

Teams evaluating Personal AI should test how the new memory layer handles deletion, export and residency requests before signing enterprise contracts.