The Quick Read
- Inclusive and Gender Responsive AI means more than removing obviously sexist outputs. It requires women’s realities, data, languages, safety and economic opportunities to shape the entire AI lifecycle.
- The India AI Governance Guidelines, released in November 2025, place fairness, equity, accountability, transparency and a people-first approach among seven governing principles.
- The India AI Impact Summit 2026 gave inclusion a prominent place through its “People, Planet and Progress” framework, the AI by HER challenge and a casebook featuring 23 gender-responsive solutions from the Global South.
- The opportunity is substantial. AI can improve access to healthcare, public services, credit, agricultural advice, safety tools and employment.
- The risks are equally real. AI can reproduce gender bias in hiring, healthcare and lending, intensify deepfake abuse and disrupt occupations where women are highly concentrated.
- India needs gender audits, representative data, women in AI leadership, accessible grievance systems, inclusive procurement and skilling linked to real jobs.
- An AI system should not be called inclusive merely because women can access it. Women must also influence what it is built to do, how it is tested and who is accountable when it causes harm.
Inclusive and gender responsive AI
A woman applies for a job. An AI system screens her CV, compares her career history with previous successful employees and quietly lowers her score. Her employment break resembles the patterns the system has learnt to associate with weaker progression.
- A woman describes chest discomfort to a health tool. Its model has been trained on data that reflects men more fully, so the risk is underestimated.
- A rural entrepreneur tries to use an AI assistant. It technically supports her language, but not her dialect, speech pattern or low-connectivity environment.
- A young woman finds that an ordinary photograph of her has been turned into sexually explicit synthetic content.
These look like separate problems. They share one question:
Whose life did the AI system understand when it learnt how the world works?
India now sees artificial intelligence as a foundation of economic growth, public-service delivery and its ambition to become Viksit Bharat by 2047. The country’s policy vocabulary is increasingly built around safe, trusted and inclusive AI.
That ambition is welcome. Yet inclusive and gender responsive AI cannot remain a statement about technology benefiting “everyone”. Whenever a society is unequal, a system built from its data can learn those inequalities, automate them and distribute them at enormous scale.
The central challenge is not whether AI will include women after it has been developed. It is whether women’s rights, work, safety, knowledge and lived realities will shape the technology from the beginning.
What is Inclusive and Gender Responsive AI?
Inclusive AI is designed so that people across languages, regions, incomes, ages, abilities and social identities can use it meaningfully and share in its benefits. Gender-responsive AI goes further. It examines how gender affects:
- The data used to train a system;
- The problem selected for AI intervention;
- The people designing and financing the technology;
- Access to devices, connectivity and digital skills;
- How an algorithm performs for different groups;
- Exposure to errors and abuse;
- The distribution of jobs, income and productivity gains;
- Who can question or appeal an automated decision.
A gender-responsive system does not simply avoid producing a stereotypical sentence about women.
- It asks whether women are represented adequately in the underlying data.
- It checks whether the product works for women with different ages, castes, disabilities, incomes, languages and locations.
- It anticipates gendered forms of misuse. It creates a usable route to remedy when something goes wrong.
UN Women describes gender-responsive AI as technology that is inclusive, equitable, safe and trustworthy, while advancing rather than undermining gender equality. Its approach places gender considerations throughout design, deployment, governance and evaluation.
This distinction matters because an AI application can be technically impressive and socially exclusionary at the same time.
India has already placed inclusion inside its AI vision
The India AI Governance Guidelines were released by the Ministry of Electronics and Information Technology in November 2025 under the IndiaAI Mission.
They adopt an agile, innovation-oriented governance model rather than proposing an immediate, comprehensive AI law.
The 7 Principles
The framework is organised around seven principles:
- Trust as the foundation;
- People first;
- Innovation over restraint;
- Fairness and equity;
- Accountability;
- Understandable by design;
- Safety, resilience and sustainability.
The guidelines also propose an institutional architecture involving an AI governance body, technical and policy expertise, safety testing and stronger coordination across sectors. Their central philosophy is that India should encourage innovation while using existing laws, standards, technical safeguards and targeted interventions to address risk.
That approach offers room for experimentation. It also places substantial responsibility on the quality of implementation.
“Fairness and equity” can remain a principle unless agencies and companies are required to explain:
- Which groups were considered;
- What data gaps were found;
- How outcomes differed by gender;
- What level of disparity was accepted;
- Who authorised deployment;
- What remedy is available.
The India AI Impact Summit 2026 continued the same inclusive language. Held in New Delhi in February, it was organised around the three pillars of People, Planet and Progress and seven thematic areas, including Inclusion for Social Empowerment, Safe and Trusted AI and Human Capital.
The summit also featured the AI by HER challenge and the launch of a casebook on AI and gender empowerment.
Produced by the IndiaAI Mission and UN Women, with support from the Ministry of Women and Child Development, the casebook presented 23 real-world solutions selected from 235 submissions across more than 50 countries. The examples showed how AI could support gender equality in areas such as safety, health, livelihoods and access to services.
India has therefore moved beyond ignoring gender in its public AI discourse. The next step is to turn attention into enforceable practice.
Why can apparently neutral AI produce gendered outcomes?
AI systems learn patterns. Those patterns come from data created by institutions, markets and people. If the historical world was unequal, the data will contain evidence of that inequality.
- A hiring model trained on past leadership appointments may learn that senior leaders are usually men.
- A lending model may interpret lower asset ownership or interrupted earnings as higher risk, without recognising the gendered structures that produced them.
- A language model trained on large volumes of online content may absorb stereotypes associating women with domestic roles and men with authority.
UNESCO tested prominent large language models and found clear evidence of regressive gender stereotyping. In one model, women were described in domestic roles four times as often as men. Female names were more frequently linked with home, family and children, while male names were associated with business, executive roles, salary and career.
The system did not invent those associations from nothing. It learnt them from society. But once embedded inside an AI tool, an old stereotype can become faster, less visible and more difficult to challenge.
A human recruiter may express bias in one decision. An automated system can repeat a similar pattern across thousands of applications while appearing objective. That is why equality cannot be measured by whether the algorithm was explicitly told to discriminate.
It must be measured through what the algorithm actually does.
Data inclusion is not a matter of adding more women to a spreadsheet
The common solution to bias is “better data”. That is correct but incomplete. Data may include women and still fail to represent them properly.
Consider a healthcare dataset containing women primarily during pregnancy. It includes women, but it may not capture their cardiovascular, neurological, occupational or ageing-related health sufficiently.
A financial dataset may contain women borrowers but too few women from rural, low-income or informal-work backgrounds.
A voice system may work for urban women speaking standard Hindi or English while failing for older women, regional accents or people with speech disabilities.
Representative data requires attention to differences among women, not only between women and men. India’s diversity makes this especially important.
Gender intersects with:
- Caste and tribe;
- Disability;
- Income;
- Rural or urban location;
- Age;
- Literacy;
- Language;
- Migration;
- Occupation;
- Access to devices;
- Family and care responsibilities.
A model can perform well for educated metropolitan women and still fail the women most dependent on public services. Gender-disaggregated data is the starting point. Intersectional testing is the standard India should aim for.
Inclusion must begin with the problem AI is asked to solve
Bias does not enter only through data. It enters through problem selection.
- A company may invest in AI to monitor employee productivity but not to identify patterns of harassment, pay inequality or women’s attrition.
- A city may build an advanced traffic-management system without asking whether women feel safe walking from the final transport stop.
- A bank may automate fraud detection without building an accessible process for women whose accounts are controlled or misused by family members.
- A public-health system may optimise patient throughput while overlooking the additional time women spend arranging care for everyone else.
The choice of problem reflects power. People build technology around the needs they notice.
When women are missing from product leadership, policy design and investment decisions, many of their problems appear secondary until a system fails.
Inclusive AI therefore requires women not only as users, coders or data subjects. Women must participate in deciding what deserves to be built.
Women are under-represented in the AI rooms that matter
Women’s participation in AI skills is improving, but leadership remains uneven.
LinkedIn and World Economic Forum data found that women’s share among people listing AI engineering skills rose from 23.5% in 2018 to 29.4% in 2025. The gap narrowed in 74 of the 75 economies studied. That is meaningful progress. It still means women form less than one-third of the visible AI engineering talent measured.
The gap becomes sharper at senior levels. Women form just over 28% of the broader STEM workforce globally and only about 12% of STEM executives, according to World Economic Forum and LinkedIn analysis.
India faces a similar pipeline challenge. Change in Content has previously examined how women remain under-represented in AI roles even as they are deeply affected by AI adoption.
Representation matters for more than fairness in hiring. Teams with limited lived experience are more likely to overlook:
- Gendered safety risks;
- Reproductive and women’s health gaps;
- Care-related career interruptions;
- Language and literacy barriers;
- Image-based abuse;
- Domestic control of digital devices;
- The difference between household access and women’s independent access.
A woman engineer does not automatically produce a gender-responsive product. A team that excludes women makes it easier for gendered blind spots to survive.
Access to AI will not be equal merely because the tool is available
India’s scale creates a powerful opportunity for AI-enabled public services.
The Government has highlighted platforms such as BHASHINI, which supports dozens of text and voice languages, and Kisan e-Mitra, a voice-based agricultural chatbot operating in regional languages. These show how language and voice interfaces can extend digital services beyond English-speaking users.
Yet availability is not the same as usable access. Women may have:
- Less independent control over a phone;
- Less private time online;
- Lower confidence in unfamiliar digital services;
- Limited data or connectivity;
- Greater exposure to fraud and harassment;
- Lower access to formal AI training;
- Fear that AI use will make their competence appear weaker.
Change in Content has reported that women’s adoption of generative AI tools remains below men’s, despite the growing importance of AI at work. The women-and-AI adoption gap cannot be explained only through confidence or technical interest.
There can also be a social penalty. In one recent experiment, identical AI-assisted CVs were judged differently when assigned male and female names. Reviewers were more likely to question the woman’s competence and trustworthiness. Our report on the gender penalty for using AI examined how equal access can coexist with unequal permission to use the tool.
That changes the policy question. It is not enough to ask whether women can use AI. We must ask what happens to women when they do.
AI can improve women’s lives in practical ways
A gender-responsive approach should not become a catalogue of risk. AI offers substantial opportunities when designed around real needs.
Healthcare
AI can support earlier identification of health risks, improve diagnostic assistance, translate health information and connect women with confidential guidance. It can be valuable where specialists are scarce, or women face social barriers in seeking care.
The system must still be clinically validated across women’s bodies, ages and health conditions. An application that expands access but produces uneven medical accuracy can create a new form of exclusion.
Livelihoods and entrepreneurship
AI tools can help women:
- Translate product descriptions;
- Create marketing material;
- Understand market prices;
- Manage accounts;
- Reach customers;
- Complete formal documentation;
- Receive agricultural or business guidance.
These capabilities can reduce some of the resource disadvantages faced by small entrepreneurs.
The tools must work through local languages, voice interfaces, low-bandwidth connections and affordable devices. Otherwise, they mainly increase the productivity of people who were already digitally advantaged.
Public services
AI can help citizens navigate complex welfare systems, identify eligibility and submit requests.
A gender-responsive public assistant should account for women whose documents, phones or bank accounts may not be under their independent control. It should also provide a human escalation route when the automated system fails.
Safety and access to justice
AI can detect abusive content, identify coordinated harassment, support evidence preservation and help users locate legal or crisis services. The IndiaAI–UN Women casebook includes examples of AI being used to address gendered harm and expand access to support.
These tools should support survivors rather than place them under greater surveillance.
Education and skilling
AI tutors can personalise learning and help women resume education or training around work and care responsibilities. However, an AI course certificate has limited value unless women gain devices, practice time, industry exposure and pathways into paid roles.
Skilling must connect with economic power.
AI can also magnify violence against women
Generative AI has lowered the effort required to produce convincing manipulated images, audio and text.
Women are already a major target of non-consensual intimate imagery, impersonation, sexualised deepfakes and coordinated disinformation.
Change in Content’s analysis of AI-enabled violence against women and girls showed how these harms can affect employment, reputation, political participation and mental health.
The damage does not stay online.
A fabricated image can reach a woman’s family, employer or community. A fake audio recording can affect an election or public career. Repeated abuse can force women to leave digital spaces that are increasingly essential for work and civic life.
Gender-responsive AI governance must therefore address both the creators of harmful content and the platforms through which it spreads.
Necessary measures include:
- Reliable provenance and watermarking tools;
- Fast removal procedures;
- Preservation of evidence;
- Clear liability and reporting channels;
- Survivor-centred human review;
- Penalties for repeat abuse;
- Support that does not require the victim to prove every technical detail;
- Special protection for journalists, public figures, minors and other high-risk groups.
Safety tools should not become another burden handed to women. Companies creating powerful generative systems must anticipate foreseeable gendered misuse.
Women may face greater disruption in the AI labour transition
The impact of AI on work will not be evenly distributed.
Women are heavily represented in clerical, administrative, customer-service and process-oriented occupations. These roles are highly exposed to generative AI because many of their tasks involve producing, organising and processing information.
The International Labour Organisation’s 2025 global index found that women’s employment is more concentrated in occupations with higher GenAI exposure. Subsequent ILO data reported that around 29% of female-dominated occupations were exposed to GenAI, compared with 16% of other occupations.
Exposure does not mean every job will disappear. Many roles will be transformed rather than fully automated. The risk lies in who receives the new work.
A company may automate routine administration and create higher-value roles involving AI supervision, analysis, client judgement and system design. If women lose the first category while men receive more of the second, technology will widen an old occupational divide.
Our analysis of AI and employment risk for women argued that transition planning must come before displacement.
That means:
- Identifying exposed roles by gender;
- Providing paid learning time;
- Moving women into augmented roles;
- Protecting wages during transition;
- Offering technical and commercial assignments;
- Measuring who receives productivity gains and promotions;
- Preventing AI adoption from becoming an easy route to remove junior women.
Viksit Bharat cannot be built by improving national productivity while weakening women’s economic participation.
A people-first framework needs people who can challenge the machine
The India AI Governance Guidelines place accountability and understandable design among their core principles.
These principles become important when AI influences consequential decisions involving:
- Employment;
- Credit;
- Insurance;
- Education;
- Healthcare;
- Welfare eligibility;
- Policing;
- Access to public services.
A person affected by such a system should be able to know:
- That AI was used;
- What role it played;
- Which information influenced the result;
- How to correct inaccurate data;
- How to request human review;
- Who is legally responsible.
“Computer says no” cannot become an acceptable administrative answer.
It is particularly important for women with limited digital literacy or institutional power. A technically available grievance portal may be unusable for someone without independent device access, formal documentation or confidence in written English.
Accountability must be designed around the person most likely to struggle with the system, not the person most comfortable with it.
What India must build into its AI governance architecture
India does not need to attach the word “gender” to every AI programme. It needs gender-responsive requirements within the systems that decide what can be deployed.
1. Mandatory gender-impact assessments for high-risk AI
Before an AI system is used in hiring, lending, healthcare, education, welfare or policing, its likely effects on women and other groups should be assessed.
The assessment should examine:
- Data representation;
- Error rates;
- False positives and negatives;
- Accessibility;
- Foreseeable misuse;
- Care and safety implications;
- The remedy available after harm.
The results should be documented rather than left to informal product discussions.
2. Disaggregated testing and public reporting
A system’s average accuracy can hide significant gaps. Companies and agencies should test performance across gender and other relevant characteristics. Where lawful and appropriate, results should be published in an understandable form.
A 92% average accuracy rate offers little reassurance when one group experiences substantially more errors.
3. Gender expertise inside AI review bodies
Technical experts are essential. So are labour economists, women’s rights specialists, disability experts, social scientists, healthcare professionals, and people from affected communities.
AI governance cannot be left entirely to engineers and lawyers. Social impact is not a technical footnote.
4. Women in leadership across the AI value chain
India needs more women as:
- AI researchers;
- Data scientists;
- Product leaders;
- Founders;
- Investors;
- Public-policy officials;
- Safety evaluators;
- Procurement decision-makers;
- Board members;
- Domain experts.
Scholarships and fellowships are useful. Women must also receive access to compute resources, research funding, patents, procurement opportunities, and commercial leadership.
5. Gender-responsive public procurement
Government is one of the most powerful buyers of technology. Procurement contracts can require suppliers to disclose:
- Data sources;
- Performance differences;
- Testing methods;
- Accessibility provisions;
- Incident-reporting procedures;
- Human-review mechanisms;
- Subcontracted data and content labour.
A vendor claiming fairness should be able to show evidence.
6. A national AI incident repository with gender categories
India’s governance framework supports stronger incident monitoring. A national repository should record patterns involving:
- Discriminatory decisions;
- Deepfake abuse;
- Stalking and impersonation;
- Healthcare errors;
- Financial exclusion;
- Workplace surveillance;
- Harmful language outputs;
- Failures affecting regional-language users.
Anonymised data can help regulators and developers identify recurring problems before they become normalised.
7. Paid and practical AI skilling for women
General digital-awareness workshops will not be enough. Women need role-specific training that helps them use AI in:
- Farming;
- Teaching;
- Healthcare;
- Manufacturing;
- Administration;
- Creative work;
- Small businesses;
- Professional services.
Training should include verification, privacy, security and the limits of AI.
Employers must provide access during working hours. Women should not be expected to learn the technology privately after completing paid and unpaid work.
8. Stronger protection against AI-enabled gender-based harm
Reporting and removal procedures should be rapid, multilingual and survivor-centred.
Police, courts, employers, platforms and educational institutions need clear protocols for synthetic sexual content, impersonation and AI-enabled harassment.
The victim should not be told that nothing can be done because the image is “not real”. The harm is real.
What must companies do now?
Organisations do not need to wait for every rule to be finalised. Any business using AI should create an internal register answering:
- Which systems are in use?
- What decisions do they influence?
- Which employees or customers are affected?
- Who approved them?
- What data is used?
- Has performance been tested by gender?
- Can a human overturn the result?
- How are complaints recorded?
- When will the system be reviewed?
Companies should also examine workplace adoption.
- Who receives premium AI tools?
- Who is invited to pilot them?
- Who gets training?
- Whose productivity is measured more aggressively?
- Who receives the higher-value work after automation?
- Are women penalised for disclosing AI use while men are praised for efficiency?
AI governance is not only a question for technology teams. It is a leadership, HR, procurement, legal and business issue.
What does inclusive AI look like in practice?
Imagine a government AI assistant helping citizens apply for a business-support programme. A technically functional version might:
- Answer questions;
- Identify eligibility;
- List required documents;
- Direct the applicant to a portal.
A gender-responsive version would also consider that some women:
- Share a phone;
- Cannot read lengthy text;
- Speak a regional dialect;
- Have limited control over family financial records;
- Run informal businesses;
- Need privacy;
- Cannot travel repeatedly to an office;
- May have names recorded differently across documents.
It would provide voice access, regional-language support, simple explanations, assisted-service options, human escalation and clear privacy protections. It might also analyse who begins and completes the application, where women drop out and why.
The technology is not gender-responsive because the home screen shows a woman. It is gender-responsive because the service has been redesigned around how different women actually encounter it.
Inclusive AI is also a question of economic ownership
Much of the public debate focuses on preventing harm.
- India should also ask who will own the value AI creates.
- Will women-led companies receive investment?
- Will women researchers own intellectual property?
- Will female workers share productivity gains?
- Will women-owned MSMEs gain procurement opportunities?
- Will women move into AI-enhanced roles or remain providers of low-paid data work?
AI supply chains already depend on large amounts of human labour, including data labelling, content moderation, testing and verification. These roles can be invisible and poorly paid.
A country cannot claim gender-responsive AI while women absorb harmful content, clean datasets and perform repetitive digital labour without adequate protection or career mobility.
Inclusion should reach ownership, income and authority.
The choice is larger than avoiding biased algorithms
The phrase “AI bias” can make the issue sound like a faulty technical output that engineers can repair. The stakes are larger.
AI will increasingly influence:
- Whose application is seen;
- Whose symptoms are taken seriously;
- Who receives credit;
- Which language is supported;
- Whose face is represented;
- Which job is redesigned;
- Who is monitored;
- Who can appeal;
- Who profits.
These are decisions about power. India’s people-first AI ambition will be credible when people who have historically held less power can shape those decisions.
That requires women inside the laboratory, the start-up, the ministry, the standards committee, the investment meeting and the user-testing room. It also requires listening to women who may never describe themselves as working in technology but will live with its consequences.
The Change Ahead
India’s AI moment is being framed as a national development opportunity. It can be one.
AI can help overcome shortages, translate services into more languages, support health workers, improve education, help small enterprises and extend public systems to people who have been difficult to reach.
But scale magnifies design choices. A biased manual process may harm hundreds. A biased automated system may harm millions before the pattern becomes visible.
The India AI Governance Guidelines have already placed fairness, accountability and people-first development inside the national framework. The India AI Impact Summit has made inclusion part of India’s global AI message. The real measure comes next.
- Will gender-impact testing become routine?
- Will women influence AI investment and product design?
- Will public systems provide human appeal?
- Will companies prepare women for changing jobs?
- Will deepfake abuse produce swift consequences?
- Will local-language AI work for the women most dependent on it?
- Will the economic value of AI be shared more fairly?
Inclusive and Gender Responsive AI does not ask technology to solve gender inequality on its own. It asks technology not to deepen inequality while claiming to modernise the country.
India cannot reach Viksit Bharat by building intelligent systems that understand only a partial version of its people. The future must recognise women accurately. More importantly, women must have the power to help write it.
Frequently Asked Questions
Q: What is Inclusive and Gender Responsive AI?
A: Inclusive and Gender Responsive AI is designed, tested and governed to work fairly for people across genders and social identities. It considers women’s data, access, safety, employment and lived experiences throughout the AI lifecycle.
Q: Why is gender-responsive AI important for India?
A: AI will increasingly influence jobs, public services, healthcare, credit and education in India. Without gender-responsive design, existing inequalities can be reproduced at national scale and weaken women’s participation in Viksit Bharat.
Q: How can AI discriminate against women?
A: AI can reproduce bias through unrepresentative data, historical employment patterns, stereotypes and unequal access. This can affect hiring, lending, medical decisions, workplace evaluation and online safety.
Q: What do the India AI Governance Guidelines say about inclusion?
A: The guidelines include people-first development, fairness and equity, accountability, understandable design, safety and trust among their central principles. Their effect will depend on how these principles are translated into testing, oversight and remedy.
Q: Can AI also empower women?
A: Yes. AI can improve access to healthcare, business tools, education, agricultural advice, public services and safety support. Its benefits are strongest when tools are affordable, multilingual, accessible and designed around women’s actual circumstances.
Q: What should companies do to make AI gender-responsive?
A: Companies should test outcomes by gender, review training data, include women in product leadership, assess job impacts, create human appeals and publish clear accountability for high-impact AI decisions.
Editorial Note and Sources
This editorial is based on the India AI Governance Guidelines released in November 2025, official material from the India AI Impact Summit 2026, international research on AI and gender and Change in Content’s previous reporting. The term “gender-responsive AI” refers to an approach to design and governance rather than a single technical standard. AI outcomes differ according to sector, model, data and deployment context. The examples in this article illustrate recognised risks and opportunities; they should not be interpreted as evidence that every AI system produces the same effects.
Principal sources
- Ministry of Electronics and Information Technology: India AI Governance Guidelines, November 2025.
- IndiaAI: Official overview of the India AI Governance Guidelines.
- Press Information Bureau: India AI Impact Summit 2026 and its People, Planet and Progress framework.
- Press Information Bureau: Seven Chakras of the India AI Impact Summit 2026.
- IndiaAI Mission, UN Women and Ministry of Women and Child Development: Real-World Impact of AI and Gender Empowerment.
- UN Women: Guidance on partnerships for gender-responsive artificial intelligence.
- UNESCO: Research on bias against women and girls in large language models.
- International Labour Organisation: Generative AI and Jobs: A Refined Global Index of Occupational Exposure.
- World Economic Forum and LinkedIn: Gender Parity in the Intelligent Age.
