Sovereign AI, male-dominated by design: how EU procurement ignores women-led tech
Women face greater exposure to AI taking over their tasks, are less represented among the people who build the systems, and are under-financed as founders.
In high-income countries, 9.6 percent of employed women work in jobs in the highest exposure category for generative artificial intelligence (AI), against 3.5 percent of men.
Women make up 19.5 percent of information and communication technology specialists in the EU. Start-ups with at least one woman founder receive 12 percent of venture capital .
Women face greater exposure to AI taking over their tasks, are less represented among the people who build the systems, and are under-financed as founders.
Our task force within the Connecting Women in Digital Forum asked a narrow question: where does AI policy risk leaving women out, and which instruments already in motion could change that?
Nobody has defined a good AI job. The EUCommission could do so, across wages, security, autonomy, progression and voice, in the Quality Jobs Act due later this year.
Member states also have until 2 December to transpose the Platform Work Directive , whose safeguards on algorithmic management are a ready model they can choose to extend beyond platform work.
Transition support follows territories, while AI exposure follows tasks. An administrative worker in a region with no designated just transition territory gets nothing. The next EU budget could make occupational exposure an eligibility trigger, with recipients reported by sex.
European support for women founders stops at a €75,000 grant . Nothing covers the step from there to €2m. The Commission’s Startup and Scaleup Strategy does not even name that gap; the only one it acknowledges sits above €100m. The Action Plan for women in research, innovation and start-ups, still to be adopted, should include a dedicated facility.
Digital Decade reporting counts adoption, not who benefits. Every AI indicator should be paired with a result reported by sex, starting with the December revision of the national roadmaps.
The fifth gap is procurement and it sits at the centre of a debate the EU is already having.
Europe's conversation about technological sovereignty usually begins with infrastructure: compute, chips, cloud and data centres. Sovereignty is also produced through purchasing: public contracts decide which suppliers gain revenue, references, scale and credibility.
In concentrated AI and cloud markets, one significant public contract is often what qualifies a firm to compete for the next.
Public procurement accounts for around 14 percent of EU gross domestic product. The commission's proposal for a Public Procurement Act, tabled on 9 September, treats that spending explicitly as strategy: simpler rules, but also environmental, social, innovation, security and resilience considerations, and a European preference.
It tracks awards to small and medium-sized enterprises, but it does not track awards to women-led firms.
The Cloud and AI Development Act , tabled in June, repeats the pattern: Article 33 requires member states to monitor cloud and AI procurement for barriers facing smaller firms and to report annually on their share of contracts, but it says nothing about gender.
Adding women-led firms to that reporting requirement would begin to show who is actually winning Europe’s cloud and AI business.
Procurement data is not currently broken down by technology category or supplier ownership, so policymakers can count contract value without knowing who got the opportunity.
In one version of sovereignty, Europe could spend more public money on AI and end up reinforcing a concentrated supplier market. In another version, public demand could deliberately broaden the range of firms that can build and scale technology in Europe.
Both count as European procurement in the statistics, yet they do not produce the same market.
Importantly, no one needs to wait for the new Act. Article 67 permits contracting authorities to use award criteria linked to the subject matter of the contract, while Article 70 allows conditions governing contract performance. That means a contracting authority can, for example, assess the inclusiveness of the team delivering a contract, where that criterion is sufficiently linked to the contract itself, or require prime contractors to report the share of subcontracted value going to women-led firms.
Spain has legislated equality criteria across every phase of procurement in its science system, and the European Institute for Gender Equality publishes a methodology . The new Act should add a duty to report awards to women-led firms by number and value, so policymakers can see whether those options are actually being used.
Each gap we identified can be addressed through an instrument already in force or on the table.
What is missing is not another framework, but action and political will power. So our ask to EU institutions, and to the governments now transposing, budgeting and negotiating these files, is simple: pick one of the five gaps, name the instrument that can close it, and name who in your institution can act on it immediately.
Catherine Mulligan leads Task Force 3 of Thematic Working Group 6 in the Connecting Women in Digital Forum. She is a visiting academic in the department of computing at Imperial College London, a Richard von Weizsäcker fellow at the Robert Bosch Academy, and a former member of the UN Secretary-General's High-Level Panel on Digital Cooperation.
Caroline De Cock is head of research at information labs, a senior fellow at the Lisbon Council and an expert member of Thematic Working Group 6, Women Shaping the AI Economy, in the Connecting Women in Digital Forum.
