top of page

Insights

The law, in plain English.

Short guides to the duties organisations already have.

Duties track

  1. 30 October: all reasonable steps to prevent sexual harassment. What changes on 30 Oct 2026, including harassment by customers and clients. What evidence of your steps can look like. (EqA s.40A; ERA 2025 ss.20–21; SI 2026/1065 reg 3)
  2. Customers, clients, contractors: the third-party duty. Who counts as a third party. What must not permit means. What to log. (EqA s.40(1A)–(1C))
  3. Evidencing your steps: a practical file. Risk assessment, survey, complaint logs, training records, review dates. Why outcomes across protected characteristics, including the sex of both parties, matter. (EqA s.40A; s.109(4))
  4. The public sector equality duty (s149), explained. Due regard to three aims, before and when decisions are made. Not box-ticking. The principles from Bracking and Bridges, stated neutrally. (EqA s.149)
  5. Equality information and objectives that measure something. The annual information and four-yearly objectives duties. Turning themes into baselines and dates. (SI 2017/353)
  6. OfS condition E6: what universities need to show. The single comprehensive source, staff–student relationships, capacity and resources. Fair process for reporting and responding students. (OfS E6, in force 1 Aug 2025)
  7. Equality action plans: getting ready for 2027. Expected for employers with 250+ staff, subject to regulations. Picking actions your data supports. (EqA s.78A; ERA 2025 s.33)
  8. Beyond the pay gap number. What gender pay gap reporting shows and what it doesn't. How equal pay for equal work differs. (EqA s.78; SI 2017/172; ss.64–66)
  9. Fathers, mothers and carers: outcomes after leave. Paternity leave, carer's leave, flexible working, redundancy protection. Tracking what happens after leave, for everyone. (Carer's Leave Act 2023; Flexible Working Act 2023; Protection from Redundancy Act 2023; EqA s.18)
  10. Reasonable adjustments: why a request log is your best evidence. The three requirements and the anticipatory duty in services. Time-to-decision and outcomes as evidence. (EqA ss.20–21, Schs 2, 8, 13)
  11. Positive action: know before you act. Allowed, not required. It needs evidence of disadvantage first. Why quotas cross the line. (EqA ss.158–159)
  12. Equality data and UK GDPR. The Art.9 condition for equality monitoring, DPIAs, and the no-group-under-10 rule. (UK GDPR Art.9; DPA 2018 Sch.1 para 8)

General information only. Not legal advice. Take advice on your own situation.

From Sentient OS: insight from our equality data analytics and our AI innovation work, side by side.

Technology track (purpose and benefit only)

  • T1. Why measuring outcomes across every protected characteristic needs new AI architecture. AI trained on past decisions inherits past gaps. Fairness has to be structural, and measured both ways. (EqA ss.13, 19)
  • T2. Fairness by design: from reporting gaps to preventing them. Reports look back. Design looks forward. Accountability by design in plain English. (EqA s.149, due regard before decisions)
  • T3. Privacy-protecting equality data. Why aggregate, minimum data and no group under 10 make evidence stronger, not weaker. (UK GDPR Art.9)
  • T4. AI in hiring: what the Equality Act already expects. Existing law applies to automated tools. No new regime is needed to ask the right questions. (EqA ss.13, 19, 39)
  • T5. Explainable decisions for boards. A board should be able to follow any decision about people. Benefits only. (EqA s.109(4))
  • T6. Why we protect our architecture as IP. Building a long-term portfolio. What patent pending does and doesn't mean. (Patents Act 1977 s.110)
bottom of page