Many European firms skipped AI over fear of unclear legal consequences, though most everyday uses are not high-risk under the EU AI Act, a Swedish vendor says.
Many European businesses that looked at AI and walked away did so because they feared the rules, not because the tools were useless. Among EU firms that considered AI and did not adopt it, 52.5 percent cite unclear legal consequences, according to official EU statistics cited by Swedish AI company Opsio. Only one in five says AI would not be useful to them.
One in five EU enterprises now uses at least one AI technology, up from 13.5 percent a year earlier. Denmark is at 42 percent, Finland at 38 and Sweden at 35. The split is by size: 55 percent of large companies use AI, against 17 percent of small ones.
Trying AI is not the problem. Generative AI use is about equal, at 37 percent of EU firms and 36 percent of US firms. The gap is in serious use. Of US businesses using AI, 81 percent apply it in more than two activities, against 55 percent in Europe.
A September 2025 survey of more than 1,000 technology small and mid-sized firms found 58 percent of EU and UK developers reporting launch delays driven by regulation. Among those delayed, half reported slower innovation and 45 percent reported higher costs.
Opsio points to a recent update. The Digital Omnibus on AI, Regulation (EU) 2026/1744, was published on 24 July 2026. It sets the dates. Rules for stand-alone high-risk AI apply from 2 December 2027. Rules for AI built into regulated products, such as machinery and medical devices, apply from 2 August 2028.
Opsio says most everyday uses are not high-risk under the EU AI Act. Examples are answering customer emails, matching invoices and sorting IT problems. The clear exceptions are AI that screens job applicants and AI that scores creditworthiness. Those are high-risk, with obligations from December 2027.
Opsio has opened an AI service to companies across the Nordic countries and the EU. It runs private, EU-hosted AI agents for customer service, finance, onboarding and IT. It says personal data is masked before any model sees it, and a person stays involved in every decision that matters. Compliance work, staff training and quarterly follow-up are included.
It starts with a free assessment. That picks three to five processes with the highest return, classifies each under the AI Act and the GDPR (the EU privacy law), and ends with a costed plan.
The announcement gives no prices and no customer results. The reading of the rules comes from the company, which sells these services, so it is worth checking against your own situation.
If you run a small business and have put AI on hold, list what you actually want it to do. Then check whether any of it involves hiring or credit decisions. If not, the rules may be less of a barrier than feared.
Before you shelve an AI project over legal worries, check whether it touches hiring or credit scoring, the clearly high-risk cases named in the story.
Nitrogen, planning software used by financial advisors, now connects to Claude, ChatGPT, Slack, Teams and Copilot, so your advisor's AI may touch your client data.
Read more: Financial advisor software Nitrogen now works inside Claude and ChatGPTGoldman Sachs reportedly expects Anthropic's sales to jump about 130 times by 2030, which could change what AI tools cost the businesses that use them.
Read more: Goldman Sachs expects Anthropic sales to hit $600 billion by 2030AWS published a free setup that gathers every AI request from all your accounts into one searchable table, so you can see which prompts and teams drive the bill.
Read more: AWS shows how to see inside every AI call your company makes
