![AU Targets AI Abuse of Women: 5 Instruments [2026]](https://shattered.io/wp-content/uploads/2026/10/african-union-ai-technology-facilitated-violence-women-peace-security-2026-gen.webp)
The African Union wants governments, AI developers and peace-and-security bodies to treat AI-enabled abuse of women as a security threat, not a side issue for content moderators. In Peace and Security Council document PSC/MIN/COMM.1339, the Council voices concern about the misuse of artificial-intelligence technologies, including cybercrime, disinformation, privacy violations, technology-facilitated violence and the potential weaponization of AI, particularly in conflict settings. That wording puts the African Union AI technology-facilitated violence debate inside the continent’s core security machinery.
A second track runs through the Office of the Special Envoy on Women, Peace and Security of the African Union Commission. The Office announced a dialogue on AI and technology-facilitated violence against women during #WPSWeek2026 in New York. Its stated themes: “Advancing the #WPSAgenda through Prevention, Protection, Responsible Technology and Women’s Leadership.” Together, the Council text and the New York dialogue show the Union moving from general AI principles toward gender-specific protections.
Other institutions are circling the same problem from different angles. Anthropic pressed courts to block risky models in its Oct. 15 letter, and UK regulators keep testing how much data they can compel from platforms. The African Union’s contribution is different. It frames abuse by AI as a threat to women’s participation in peace processes, which raises the stakes beyond individual harm. This analysis lays out what the record confirms, what it leaves open, and what builders and policymakers should watch next.
Don't miss new tech stories on Google
Add Shattered once in the Google app and our stories show up in your news suggestions.
What the Verified Record Confirms About the AU Call #
Precision matters here, because early coverage of any policy story tends to blur documents, announcements and speculation. Five items are confirmed in the public record we reviewed. Several others, often assumed in social posts, are not. The table separates them so readers can see where the evidence stops.
| Item | Status | Source basis |
|---|---|---|
| Council concern over AI misuse (cybercrime, disinformation, privacy violations, technology-facilitated violence, weaponization) | Confirmed | PSC/MIN/COMM.1339 |
| Call to build African capacity in electricity access, computing power, datasets, infrastructure and expertise | Confirmed | PSC/MIN/COMM.1339 |
| Continental AI Strategy addresses safety and security, including technology-facilitated gender-based violence | Confirmed | AU Continental Artificial Intelligence Strategy |
| Convention on Ending Violence against Women and Girls covers cyberspace, adopted February 2025 | Confirmed | AU Convention text |
| Special Envoy dialogue on AI and technology-facilitated violence at #WPSWeek2026 in New York | Confirmed (announced) | Office of the Special Envoy on WPS |
| Speaker list and attendance | Unconfirmed | Not published in the record |
| Event date and venue inside New York | Unconfirmed | Not published in the record |
| Direct quotes from the dialogue | Unconfirmed | Event had not been reported on when we checked |
| Named companies or AI products tied to the call | Unconfirmed | No product or firm is named in the record |
| Numerical findings on AI abuse tied to this headline | Unconfirmed | No dataset is attached to the announcement |
The last four rows deserve attention. No company is named in either the Council text or the dialogue announcement. No prevalence statistic comes attached to the call. A reader who sees a viral post naming a specific app or model in connection with this story is looking at something the record does not support.
Inside PSC/MIN/COMM.1339: The Council’s AI Concerns #
The Council issued the document after its 1339th meeting, held on 16 April 2026, according to the title on the Council’s own web page. The meeting theme was Artificial Intelligence: Governance, Peace and Security. The communiqué reads as a risk register more than a regulation. It lists cybercrime, disinformation, privacy violations and technology-facilitated violence, then adds the weaponization of AI as a potential escalation, especially where armed conflict is already under way.
Two features stand out. First, the Council groups technology-facilitated violence with the classic security risks rather than leaving it under a separate social-policy heading. Second, it sets the whole discussion in conflict settings. A deepfake that damages a community organizer in a stable city is serious. The same fake, released during a ceasefire negotiation, can derail talks. The Council’s phrasing leaves room for both readings.
The communiqué also looks inward at the continent’s own position. It calls for efforts to build African capacity in AI-related electricity access, computing power, high-quality datasets, infrastructure and technical expertise. That list is a diagnosis. African institutions, in this telling, cannot police AI misuse when they lack the power grids, compute and data to build or audit the systems themselves.
Why Peace and Security Is the Right Frame for AI Violence #
Many national laws handle online abuse as a crime between individuals. The African Union is applying a collective lens. Women’s participation in peacemaking, mediation and political life is a security asset, and attacks on it weaken the process. Harassment campaigns that push a woman negotiator offline do not only hurt her. They narrow the set of voices at the table.
That logic traces back to the Women, Peace and Security agenda, which the UN Security Council began with Resolution 1325 in October 2000, according to UN Women’s Africa office. UN Women’s October 2026 special edition argues that sustainable peace cannot be achieved without women’s full, equal and meaningful participation. Add generative AI to that picture and the threat model changes. Abuse that once needed time, skill or a large troll network can now be produced fast and cheaply.
We should be careful not to overstate what the Council has proven. The communiqué expresses concern. It does not publish incident counts. Its value lies in agenda-setting, which in the AU system often precedes budget lines, model laws and monitoring mechanisms.
The Continental AI Strategy and the 2025 Convention #
The Council’s communiqué does not stand alone. Two older instruments already give the Union legal and policy hooks for the topic. Both predate or sit alongside the April meeting, and both matter for how far the new call can travel.
African Union Continental Artificial Intelligence Strategy
The Strategy addresses safety and security and names multiple forms of technology-facilitated gender-based violence in that context. For developers, the practical point is that the Union’s flagship AI policy already treats gendered abuse as part of the safety scope, not as an afterthought. Any firm building products for African users can expect that framing to appear in procurement terms and regulatory expectations over time.
Convention on Ending Violence against Women and Girls
The African Union Convention on Ending Violence against Women and Girls was adopted in February 2025 and explicitly covers violence occurring in cyberspace. The record we reviewed does not say how many member states have ratified it, so we make no claim on that point. Adoption alone still matters. It gives advocates a continental text to cite when national legislatures write rules on synthetic sexual imagery, stalking tools and coordinated harassment.
| Instrument | Type | Relevance to AI abuse of women | Timing in the record |
|---|---|---|---|
| PSC/MIN/COMM.1339 | Council communiqué | Flags technology-facilitated violence and weaponization of AI in conflict | Meeting of 16 April 2026 |
| Continental AI Strategy | Continental policy | Covers safety and security, including technology-facilitated gender-based violence | Adoption date not verified here |
| Convention on Ending Violence against Women and Girls | Continental convention | Explicitly covers violence in cyberspace | Adopted February 2025 |
| Special Envoy dialogue | Convening | Prevention, protection, responsible technology, women’s leadership | #WPSWeek2026, New York, date unconfirmed |
| Office of the Special Envoy Strategic Plan | Institutional plan | Sets the Office’s direction for 2026 to 2028 | Launch reported 15 September 2026 |
The last row comes from the African Union’s Women, Gender and Youth Directorate news page, which lists the Strategic Plan 2026 to 2028 launch of the Office of the Special Envoy on Women, Peace and Security. The page does not say whether AI features in the plan, so we draw no link beyond the Office’s mandate.
The #WPSWeek2026 Dialogue in New York #
The Office of the Special Envoy announced the dialogue as part of #WPSWeek2026 in New York. Its four stated themes are prevention, protection, responsible technology and women’s leadership. Those four words work as a policy checklist. Prevention points at design choices and platform duties. Protection points at legal remedies and survivor support. Responsible technology points at developers and deployers. Women’s leadership points at who sits in the rooms where these rules get written.
New York as a venue matters too. The Union is taking an African position to the UN’s home turf, where member states, UN agencies and technology firms with global reach all gather. That choice suggests the Office wants the African Union AI technology-facilitated violence message to shape global WPS language, not only continental policy. What we cannot report is who spoke, who attended or what the dialogue concluded. The announcement did not name speakers, and we found no readout. Any article claiming outcomes today is guessing. We will update this analysis if the Office publishes a summary.
What the Council and Commission Chair Said, in Their Own Words #
Direct statements add texture to the policy record. The lines below come from Council communiqués and a statement by the Commission Chair. We quote them verbatim and attribute them to the bodies that issued them.
The Council’s AI communiqué, the same PSC/MIN/COMM.1339 text linked above, says it “Expresses concern over the risks associated with the misuse of AI technologies, including cybercrime, disinformation, privacy violations, technology-facilitated violence and the potential weaponization of AI, particularly in conflict settings.”
An earlier Council communiqué on women’s leadership and emerging threats (document number 1334 in the AU series) goes further on the gender dimension. It “Acknowledges that while offering opportunities for innovation, socio-economic development, inclusion of women, emerging technologies, including Artificial Intelligence and digital platforms also present new risks such as technology-facilitated gender-based violence, online harassment, misinformation, and disinformation, which undermines women’s effective leadership, credibility, reputation, participation, safety, and authority.”
The same document “Expresses deep concern over the ongoing violent conflicts in the Continent which are disproportionately affecting women and girls.” It also notes the rise of technology-facilitated violence, misinformation and disinformation, which it says constrain women’s participation in governance, peace and decision-making processes.
It then turns to member states. The Council “Calls upon Member States and AU institutions to operationalize gender-responsive policies, legal frameworks, and institutional measures to prevent technology-facilitated violence against women, promote safe digital environments, and advance ethical deployment of Artificial Intelligence and digital technologies.”
African Union Commission Chairperson Mahmoud Ali Youssouf struck a similar note in a statement on AI and its effect on peace, security and governance in Africa. He said, “Left unchecked, AI will amplify these threats, exploiting disinformation, cyber-attacks, and even autonomous weapons.”
Read together, the quotes show a consistent chain of logic. Technology creates opportunity, abuse erodes women’s participation, and states must write rules and build institutions. The word operationalize carries weight. The Council asks for working policies, not declarations.
How AI Models Enter the Abuse Chain #
The AU documents speak in categories rather than product names, so this section is our analysis, not a restatement of the record. Generative tools can lower the cost of several harms the Council lists. They can produce convincing fake video or audio of a real person. They can write persuasive harassment at scale in many languages. They can help an attacker draft phishing lures aimed at a specific organizer. Each of those maps onto the Council’s list of disinformation, privacy violations and cybercrime.
Capability evidence from the security world shows how quickly the offensive side is moving. We covered a test in which Anthropic models hacked three organizations in capture-the-flag runs. That research concerns cyber operations rather than gendered abuse, but the underlying lesson carries over. Models that can plan and execute multi-step tasks can also plan harassment or surveillance campaigns if guardrails fail.
Developers have started adding narrow blocks in response to specific misuse. One example is the flag we described in Claude Code’s block on video game recreation. Gendered abuse needs similar targeted controls, from refusals on non-consensual intimate imagery to rate limits on mass impersonation. The AU documents do not prescribe such controls. They create the political space to demand them.
One caution applies to anyone reading model-release news alongside this story. Our research pass could not tie any specific new model from the last one to three months to non-consensual imagery in a way the sources confirm. We therefore name no model as a driver of the AU concern. Linking a product to harm without evidence would repeat the speculation this analysis tries to avoid.
The Capacity Gap: Electricity, Compute and Data #
The Council’s capacity language deserves more attention than it usually gets in coverage of AI safety. Detecting synthetic abuse, auditing a model and running a national response unit all require infrastructure. If a government cannot reliably power data centers, buy accelerators or assemble local-language datasets, it depends on foreign vendors for both the tools and the judgment about what counts as abuse.
That dependence has a security angle. Open-weight models spread capability widely, which helps local builders and also helps bad actors. Our earlier report on Mistral Large 4’s open weights shows how fast frontier-scale systems reach anyone with enough hardware. The same dynamic applies to image and voice generators. Once weights circulate, a policy built around vendor cooperation loses force.
Cost is another factor. Epoch AI’s test, which we covered when GLM 5.3 cracked six Firefox builds at $1,000 each, put a price tag on automated offensive work. Cheap automation changes who can mount a campaign. For women in public roles across Africa, falling attack costs mean a larger pool of potential harassers, including those with no technical skill.
Conflict Backdrop: What the UN Women Numbers Show #
The Council places its AI concerns in conflict settings, so the wider WPS data helps frame the stakes. UN Women’s Africa office published a set of global figures in October 2026, drawing on the UN Secretary-General’s 2026 report on Women, Peace and Security (S/2026/758). None of these numbers measure AI abuse directly. They describe the environment in which that abuse lands.
| Indicator | Figure | Period |
|---|---|---|
| Women and girls living within 50 kilometers of armed conflict worldwide | 756.1 million | 2025, highest since the 1990s |
| Women among negotiators in tracked formal peace processes | 5% | 2025 |
| Peace agreements that referenced women, girls, gender or sexual violence | 19% | 2025 |
| Countries with at least one National Action Plan on WPS | 117 | As of May 2026 |
| Fall in official development assistance | 23% | 2025, largest annual drop |
| Women’s organizations surveyed with under six months of funding left | 45% | 2026 UN Women survey |
Two readings follow. First, women hold few seats in peace processes, so every attack that pushes one of them offline carries outsized weight. Five percent of negotiators is a small base. Second, the money that funds women’s organizations is shrinking. Groups that monitor online abuse, run helplines or train journalists compete for the same dollars. Policy ambition without funding will stall.
How the AU Approach Compares With Regulator-Led Models #
The African Union works mainly through norms, communiqués and conventions. Its power depends on member states to turn text into law. Compare that with regulator-led models. In the UK, Ofcom has used formal information notices to demand data from platforms, and our report on how Meta, TikTok and X lost their High Court fight with Ofcom shows what that enforcement looks like in practice. A regulator with statutory powers can compel answers. A continental body can set direction but must rely on others to compel.
Neither model is superior in every respect. Regulators move faster against named platforms and slower on cross-border harms. Continental bodies shape a shared vocabulary that national lawmakers then borrow. The African approach fits a region with many jurisdictions at varied capacity. A single convention covering cyberspace gives smaller states a template, even if enforcement stays national.
Industry self-governance sits on a third axis. Model developers publish usage policies and train refusals, but those rules reflect priorities set elsewhere and often tested mainly in English. The AU documents implicitly challenge that arrangement. They ask for African capacity in datasets and expertise so local actors can judge what protection looks like in their own languages and contexts.
Historical Context: From Resolution 1325 to the AI Era #
The Women, Peace and Security agenda began with UN Security Council Resolution 1325 in October 2000. UN Women’s October 2026 special edition marks 26 years since that adoption and says translating commitments into lasting change remains an urgent priority. The agenda has since spread through national action plans and regional frameworks, and UN Women counts 117 countries with at least one such plan as of May 2026.
The African Union’s own history on this topic runs through its gender instruments. The February 2025 Convention on Ending Violence against Women and Girls brought cyberspace explicitly into continental law-making. The Continental AI Strategy added technology-facilitated gender-based violence to the safety and security discussion. The April 2026 Council communiqué then connected both threads to the Council’s mandate on peace and security. Each step widened the frame.
Seen over that arc, the New York dialogue looks like a continuation rather than a sudden turn. The Union has been building a policy stack since at least February 2025, and the Special Envoy’s Office is now carrying it onto the global stage.
Market and Industry Impact #
No figures in the record quantify the commercial effect, so this section describes direction rather than size. Several groups should watch the policy signals closely. Trust-and-safety vendors that sell detection of synthetic media may find new demand if member states act on the Council’s call for institutional measures. Cloud and compute providers could see interest in regional capacity, since the Council names electricity access and computing power as gaps. AI developers serving African users face reputational and, eventually, regulatory pressure to test safeguards in local languages.
Security teams at media houses, election bodies and civil-society groups also have a stake. They will need identity checks, provenance tools and incident playbooks that treat harassment of staff as a security event. Our piece on zero trust for AI agents offers a useful mental model, because the same principle of verifying every actor applies to synthetic voices and faces.
Funding limits the upside. UN Women’s figures show development aid fell 23 percent in 2025 and many women’s organizations hold thin reserves. Buyers with real budgets are therefore likely to be governments, large donors and multinational platforms, not the grassroots groups closest to the harm. Vendors who ignore that gap risk selling tools no frontline group can afford. For wider context on model-level trends, see our AI and machine learning coverage.
Five Predictions for the Next 12 Months #
These are our judgments, not reported facts. Each rests on the documents above and on how the African Union AI technology-facilitated violence agenda is likely to move through continental bodies.
- Follow-up texts will cite the Convention’s cyberspace clause. Expect draft national laws and Council statements to point to the February 2025 Convention when they address synthetic abuse.
- Capacity will become a funding ask. The Council’s list of electricity, compute, datasets and expertise gives donors a ready shopping list, and proposals will follow that structure.
- Local-language safety testing will gain attention. Pressure will build on model developers to show abuse safeguards work outside English and French.
- Conflict settings will stay on the Council’s AI agenda. The weaponization language leaves room for recurring briefings on AI use in active crises.
- Funding gaps will slow delivery. With aid falling and women’s organizations thin on reserves, monitoring and survivor support will lag behind policy ambition.
What Builders and Policymakers Should Do Now #
The African Union AI technology-facilitated violence agenda has direct consequences for teams shipping AI products into African markets, and the practical steps are plain. Map your product against the Council’s risk list and note where impersonation, harassment automation or privacy violations could arise. Test refusals on intimate-imagery requests in the languages your users speak. Publish an abuse-reporting path that a non-technical user can find. Log enforcement actions so you can answer a regulator or a Council inquiry with data.
For policymakers, the Council’s own phrasing offers the best guide. Build gender-responsive policies, legal frameworks and institutional measures, then fund them. Pair any new offense with survivor support and a clear takedown route. Involve women’s organizations in drafting, since the Council ties women’s leadership to the success of the whole effort. The takeaway: policy that treats AI abuse as a peace-and-security matter needs budget lines and institutions, not only communiqués.
FAQ: African Union AI and Technology-Facilitated Violence #
What is PSC/MIN/COMM.1339?
It is the African Union Peace and Security Council document that expresses concern about misuse of AI technologies, including cybercrime, disinformation, privacy violations, technology-facilitated violence and potential weaponization of AI in conflict settings. It also calls for building African capacity in electricity access, computing power, datasets, infrastructure and expertise.
Does the African Union already have rules covering online violence against women?
Two instruments touch it. The Continental AI Strategy addresses safety and security, including technology-facilitated gender-based violence. The Convention on Ending Violence against Women and Girls, adopted in February 2025, explicitly covers violence in cyberspace. The record we reviewed does not report ratification numbers.
What is the #WPSWeek2026 dialogue?
The Office of the Special Envoy on Women, Peace and Security of the African Union Commission announced a dialogue on AI and technology-facilitated violence against women during #WPSWeek2026 in New York. Its stated themes are prevention, protection, responsible technology and women’s leadership.
Which AI companies or models does the African Union name?
None in the documents we reviewed. The Council speaks about categories of harm and about capacity, not about specific products or firms. Claims linking a named app or model to the call are unsupported by the record.
Why does the Council link this to peace and security?
Earlier Council language says technology-facilitated violence and disinformation constrain women’s participation in governance, peace and decision-making. Attacks that push women out of those processes weaken peace efforts, so the Council treats them as a security matter.
Who attended the New York dialogue and what did it conclude?
That is unconfirmed. The announcement did not name speakers or publish an attendance list, and we found no readout of outcomes. We will update this analysis if the Office releases a summary.
What can developers do today?
Test refusals on non-consensual imagery and impersonation in local languages, publish a clear abuse-reporting route, and keep enforcement logs. The Council asks states to promote safe digital environments and ethical deployment of AI, which those steps support.