Gender Equality Plan 2026–2028

1. Why we publish this plan

AI STM Learning SRL, trading as stm.ai, is a Romanian research and development company working on applied artificial intelligence across healthcare, cybersecurity, and education. Our work touches clinical data, learning materials used by children, and security decisions taken on behalf of other organisations. In each of those domains, who is in the room, whose data is represented, and whose experience is treated as the default are questions with technical consequences, not only ethical ones.

We publish this plan because gender equality in a company of our size is easy to assert and hard to evidence. A public document with named ownership, dated commitments, and measurable targets is the only version of the claim that can be checked.

We also state the starting position plainly rather than flattering it. stm.ai is a founder-led micro-enterprise. Its leadership is currently not gender-balanced, and no single hiring decision will change that. This is precisely why the commitments below are written as process obligations — things that must be done, recorded, and reviewed whatever the composition of the team happens to be in a given year.

2. Scope, period, and application

This plan covers the period 2026–2028 and applies to every person who works for or on behalf of AI STM Learning SRL — employees, founders, contractors, collaborators, and interns — and to the company's recruitment activity, progression and pay discussions, research practice, public and product communication, partner and consortium conduct, and internal standards of behaviour.

The measures are proportionate to our current scale. We are not a university or a hospital system, and we do not pretend to operate the committee structures of one. Proportionate does not mean optional: each measure below is written so that a reader can tell, at the end of a year, whether it happened.

The plan operates alongside, and does not replace, our obligations under Romanian and EU law, including Law no. 202/2002 on equal opportunities and equal treatment between women and men, Law no. 167/2020 amending the anti-discrimination framework, and the applicable provisions of the Romanian Labour Code.

3. Governance, ownership, and dedicated resources

Overall accountability for this plan rests with Cosmin Stamate, Founder and Managing Director. Accountability is not delegated; specific tasks may be.

The company commits to:

• allocating leadership time each year to implementation, data review, and the annual progress update, and treating that review as a scheduled obligation rather than a discretionary one;
• allocating working time — not goodwill outside working hours — to communicating the policy, maintaining the indicator set, and following up on actions arising from the review;
• engaging external gender-equality expertise, including through our academic and consortium partners, where a question exceeds our internal competence or where an internal reviewer would not be independent;
• attaching every equality-related decision to a named owner and a date, so that a decision can be audited rather than inferred.

4. Data, indicators, and annual monitoring

Once per year, and using the minimum data necessary, we compile and review the following indicators where they are lawfully recorded and meaningful at our size:

• composition of the workforce and of the wider collaborator base by sex/gender;
• applicant, shortlist, offer, and acceptance counts by sex/gender for each externally published role;
• representation in leadership, decision-making, and advisory roles;
• participation in equality and unconscious-bias training;
• the number, category, and resolution status of concerns raised through the route described in section 7.

Because our headcount is small, individual figures can be identifying. We therefore report indicators publicly only in aggregate and suppress any cell that would identify an individual; where suppression makes a figure meaningless, we say so rather than publishing a misleading number. Personal data collected for this purpose is processed under the GDPR on the basis of our legal obligations and legitimate interest, held for no longer than needed for the annual cycle, and handled in line with our privacy contact at privacy@stm.ai. Providing sex/gender information for monitoring is voluntary.

A short annual progress note is published and linked from this page, stating what was done, what the indicators showed, and what changed as a result.

5. Training and awareness

Every year, all staff and everyone involved in hiring, evaluation, or supervision completes awareness training on gender equality, unconscious bias in selection and evaluation, and the company's conduct expectations. People joining the company receive the same material as part of onboarding, before they take part in any hiring or review decision.

Training is scaled to our size, which for a micro-enterprise may mean a structured self-study module and a recorded discussion rather than a facilitated workshop. What is not scalable is whether it happens: completion is recorded, and non-completion is treated as an open action rather than an accepted outcome.

6. Areas of action and targets for 2026–2028

The five areas below correspond to the recommended content areas of the Horizon Europe Gender Equality Plan criterion. Each contains what we will do and how the doing will be checked.

6.1 Recruitment, selection, and career progression

We will:

• write and review externally published role descriptions for inclusive, non-gendered wording, and state salary or salary range wherever we are able to;
• define the evaluation criteria for a role before candidates are seen, and assess every candidate against those criteria only;
• avoid single-person selection decisions for any role by involving at least one additional reviewer, internal or external;
• keep progression and pay conversations structured, evidence-based, and written down.

Targets for 2026–2028:

• 100% of externally published role descriptions reviewed for inclusive wording before publication;
• 100% of open roles have written evaluation criteria recorded before the first candidate is assessed;
• applicant, shortlist, hiring, and progression data reviewed annually for bias signals, where volumes make the review meaningful;
• every person working for the company has at least one documented career discussion per year.

6.2 Balance in leadership and decision-making

We will:

• treat the current absence of gender balance in leadership as a stated risk to be addressed as the company grows, not as a neutral fact;
• apply documented, equality-minded selection criteria to any newly created leadership, advisory, or consortium-representation role;
• seek candidate pools for such roles that are not gender-homogeneous, and record the search measures taken when they nonetheless are;
• review representation in decision-making annually, including in the partnerships we join.

Targets for 2026–2028:

• gender composition of leadership, advisory, and decision-making roles reviewed and recorded annually;
• for every new leadership or advisory role created in the period, the selection criteria, the search measures used to reach a mixed candidate pool, and the outcome are documented;
• where the company nominates representatives to a project consortium or public panel, gender balance is an explicit consideration in the nomination and is minuted.

6.3 Working culture and work–life balance

We will:

• maintain flexible and remote working wherever the role allows it;
• schedule meetings, deadlines, and travel so as not to systematically disadvantage people with caregiving responsibilities, and avoid default expectations of availability outside agreed hours;
• apply parental, carer's, and family leave entitlements consistently, without penalty to progression, and support return-to-work transitions;
• run inclusive meetings and keep workload allocation an explicit, discussable topic.

Targets for 2026–2028:

• 100% of staff receive, at onboarding and at least annually thereafter, a communication covering this plan, respectful-conduct expectations, and the reporting route;
• working-culture and inclusion feedback is collected and reviewed at least once per year;
• flexible-work and leave decisions are recorded, and the basis for any refusal is written down.

6.4 The gender dimension in research and public content

This is where our plan has the most technical bite. Our peer-reviewed work spans Parkinson's disease measurement, EEG-based studies, human motor behaviour, and machine learning; our products include a cybersecurity platform (CLARA) and personalised educational material for children (LLMentor). Sex and gender are potential confounders, potential effect modifiers, and potential sources of dataset and representation bias across all of them.

We will:

• ask, at the design stage of every research or product initiative, whether sex and/or gender is relevant to the research question, the sampling frame, the data collection instrument, the analysis, the model's training data, or the interpretation of results;
• record the answer, including a reasoned negative, rather than leaving relevance to be assumed either way;
• report the sex/gender composition of study populations and training datasets where those data exist, and name the limitation where they do not;
• check generated and curated learning content, model outputs, and public communication for gender stereotyping, and review illustrative examples and personas for the same;
• remain attentive to underserved groups and access barriers in Romanian and wider European healthcare and education contexts, where women, older women in particular, are under-represented in the evidence base.

Targets for 2026–2028:

• 100% of new research protocols and product initiatives include a recorded sex/gender relevance assessment before data collection or model training begins;
• where sex/gender relevance is affirmed, the consideration is documented in the protocol and carried through to the reporting of results;
• active research outputs and public-facing product and research communication are reviewed at least annually against this section.

6.5 Preventing gender-based violence and sexual harassment

The company holds a zero-tolerance position toward gender-based violence, sexual harassment, intimidation, and retaliation against anyone who raises a concern in good faith. This applies to conduct at company premises, at client and partner sites, at conferences and events, and in written and online channels used for work.

We will:

• maintain the clear, named reporting route set out in section 7, including a route that bypasses the Managing Director;
• handle every concern confidentially, promptly, and proportionately, and record its handling;
• protect the person raising the concern from retaliation, and treat retaliation as itself a breach;
• ensure that anyone in a supervisory position understands their escalation and response duties;
• state these expectations to partners and consortium members whose staff work alongside ours.

Targets for 2026–2028:

• the reporting route and conduct expectations remain published here and are communicated internally at least annually;
• an annual awareness reminder is completed and recorded;
• every reported concern is acknowledged within five working days, documented, reviewed, and handled through the defined response path, with the aggregate count reported in the annual note.

7. Raising a concern

A concern about discrimination, harassment, or any breach of this plan may be raised in writing to c@stm.ai. It will be acknowledged within five working days.

We recognise the structural weakness of a reporting route inside a founder-led company: if the concern is about the Managing Director, reporting it to him is not a remedy. Where that is the case, or where the person raising the concern prefers not to use the internal route for any reason, the concern may be taken directly to:

• the National Council for Combating Discrimination (Consiliul Național pentru Combaterea Discriminării, CNCD), the Romanian equality body competent to receive complaints of discrimination, including on grounds of sex;
• the Labour Inspectorate (Inspecția Muncii) or the territorial labour inspectorate, for workplace conduct and employment-law matters;
• the criminal authorities, where conduct may constitute an offence.

Using an external route is never a breach of this plan, is never treated as disloyalty, and does not require internal escalation first. Nothing in this document limits any right of action a person has under Romanian or EU law.

8. Review, revision, and publication

This plan is reviewed once per year. The review asks three questions and records the answers: whether the committed actions actually happened; whether the indicators in section 4 show an equality risk, a gap, or progress; and whether resources, training, or processes need to change as a result.

The plan may be revised before the end of 2028 where organisational growth, a funding condition, a change in law, or something we learn from the annual review makes revision necessary. Each revision increments the version number and is re-approved by top management. The current version is always the one published at this address, and superseded versions are available on request from c@stm.ai.

9. Approval

This Gender Equality Plan is approved for publication and implementation by the top management of AI STM Learning SRL.

Annex A — Mapping to the Horizon Europe eligibility criterion

Horizon Europe requires applicant organisations in the relevant categories to have a Gender Equality Plan meeting four mandatory process requirements, and recommends five content areas. This annex shows where each is met in this document.

Annex B — References

• EIGE — Horizon Europe gender equality plan eligibility criterion
• EIGE — Gender Equality in Academia and Research (GEAR) tool
• European Commission — Gender equality in research and innovation
• Law no. 202/2002 on equal opportunities and equal treatment between women and men (Romania), as republished and amended.
• Law no. 167/2020 amending Government Ordinance no. 137/2000 on preventing and sanctioning all forms of discrimination (Romania).

Questions about this plan may be sent to c@stm.ai. Data-protection questions relating to the monitoring described in section 4 may be sent to privacy@stm.ai.