Belgian AI news
Short briefings on Belgian AI, startups, policy, deployments, published Mondays and Thursdays. Each item gets a Mue take on what it means if you run a small business here.
-
OpenAI open-sources Symphony, an orchestration spec for Codex coding agents
OpenAI released Symphony, an open-source specification that turns project-management tools like Linear into control planes for Codex coding agents. Every open task gets an agent, agents run continuously, and humans review the results. The reference implementation is in Elixir, but the community has already rebuilt it in TypeScript, Python, and Rust. The repo hit 15,000 GitHub stars within days of launch.
Mue take: This is the pattern we use at Mue. If you already run a task board, you can wire up coding agents without building custom infrastructure. Start small: one repo, one task type, human review on every PR.
Source: OpenAI
-
Half of Belgian secondary school students now use AI weekly for schoolwork
A Vlaamse Scholierenkoepel survey finds 50% of Belgian secondary students use AI tools at least weekly for assignments. 62% report inconsistent rules between teachers and 42% are worried about how their classmates use it. Students are asking for unified school policies and explicit AI-literacy class time.
Mue take: Your future hires already write with AI. If your onboarding does not assume that, you are behind your interns.
Source: VRT NWS
-
Proximus and Faktion publish #Interact case study on agentic AI rollout
Proximus and Antwerp-based Faktion published the case study for #Interact, an omnichannel agentic-AI platform that unifies WhatsApp, SMS/RCS, Messenger, web chat and voice with LLM-powered assistants, built on Proximus' Enco API and now expanding to enterprise customers.
Mue take: Belgian telcos are productising agents. The cost floor for "omnichannel AI" will drop fast. If you are evaluating an enterprise contact-centre tool, wait a quarter or build the lean version yourself.
Source: Faktion
-
Belgian study: gender bias "underestimated" in AI-assisted recruitment
A Belgian academic study reports 75% of recruiters use AI in hiring but only 12-17% recognise bias risks. Researchers found AI models penalise female candidates via proxy variables even when explicit gender markers are removed. Likely to feed the FOD Economy / BIPT discussion ahead of the August 2026 AI Act high-risk deadline.
Mue take: If you screen CVs with an off-the-shelf AI tool, recruitment counts as high-risk under the AI Act. You need to know what the model is doing, document it, and be ready to defend it. Do not wait until August.
Source: Brussels Times