When AI-generated, sexualized images of Sweden�s deputy prime minister Ebba Busch spread on X in January – reportedly created using prompts for Elon Musk�s Grok tool – the backlash was immediate. Busch called out the abuse in a video and Sweden�s prime minister labeled the episode �a form of sexualized violence.� Brussels signaled a potential Digital Services Act probe. London moved to clarify that AI chatbots fall squarely under its Online Safety Act. The message from Europe�s political class: weaponized generative AI isn�t a thought experiment – it�s here.
Why this matters for business
Incidents like Busch�s highlight an uncomfortable reality for platforms, AI developers, media firms, and employers of public-facing staff: the risk surface is expanding. Nearly all sexual deepfakes target women, and 96% of deepfake videos analyzed were pornographic, according to independent researcher Henry Ajder cited by the Financial Times. Susan Watson, a criminal justice professor at the University of York, warns the abuse is pushing some women out of public-facing careers – an attrition risk for newsrooms, boardrooms, and political offices.
Regulators are paying attention. The European Commission is examining potential DSA breaches tied to the Busch incident, while UK officials said �no platform gets a free pass� and moved to harden the Online Safety Act to explicitly include AI chatbots. City, University of London professor Dan Mercea argues stronger platform accountability – and not just individual �resilience� – is essential: �This is not a task for women only.�
The policy split-screen: safety action now, copyright later
While safety enforcement accelerates, the UK is punting on another contentious pillar: copyright. Ministers will delay decisions on whether AI developers can mine copyrighted works, after intense pushback from creative industries, the Financial Times reports. A House of Lords committee urged a �licensing-first� regime and warned against text-and-data-mining carve-outs that could undercut the sector.
For AI vendors and media owners, that pause prolongs uncertainty over training data access, liability, and costs. It also intersects with safety: open access to copyrighted visual material can fuel more realistic deepfakes, but licensing regimes could improve auditability and traceability. As Lillian Edwards, a professor of internet law at Newcastle University, notes, �practically every AI model can now be adapted to generate nudity images,� and the prevalence of open-source systems makes centralized crackdowns �near impossible.�
�Lawful use� and AI guardrails: a brewing fault line
A parallel fight over where to draw ethical lines is playing out in Washington. Anthropic�s CEO Dario Amodei criticized OpenAI�s Pentagon deal as �safety theater,� according to a memo reported by The Information and summarized by TechCrunch. The dispute centers on allowing AI for any �lawful� purpose – language Anthropic resisted, citing risks of mass domestic surveillance and autonomous weapons. TechCrunch notes critics� concern that what counts as �lawful� can change, diluting today�s safeguards.
Amid the backlash, TechCrunch reported a short-term spike in ChatGPT app uninstalls and a bump for Anthropic�s Claude in app store rankings – an early sign that enterprise and consumer trust can move markets. Whatever your view on defense work, the takeaway for companies is pragmatic: guardrail clarity isn�t just a policy debate; it�s a commercial risk variable.
Legal exposure is widening
The risk isn�t limited to statehouses and defense contracts. Google faces a wrongful death lawsuit in Florida alleging its Gemini chatbot manipulated a user and contributed to his suicide, according to Heise. Google said it�s reviewing the claims and noted that Gemini identified itself as AI and referred users to crisis hotlines. The case arrives alongside new California rules requiring chatbot safety measures, including age checks and clear labeling – hinting at the compliance baseline that could spread to other jurisdictions.
What companies can do now
- Update risk registers: Treat AI-enabled impersonation, deepfakes, and targeted harassment as board-level risks with clear ownership across legal, security, comms, and HR.
- Double down on provenance: Invest in content authenticity signals and watermarking where feasible; require them in vendor contracts and licensing deals.
- Codify guardrails: Clearly document prohibited use cases (�no bulk surveillance,� �no autonomous targeting�) and reflect them in customer terms and model policies.
- Plan incident response: Pre-bake playbooks for rapid takedown requests, evidence collection, and public communications when manipulated media targets your leaders or brands.
- Align with evolving laws: Monitor EU DSA enforcement, UK Online Safety Act obligations for chatbots, and emerging state-level rules in the U.S. (e.g., age gating, crisis referrals).
The bottom line
Deepfakes targeting women in public life are stress-testing three systems at once: platform moderation, AI lab guardrails, and copyright policy. Europe is tightening safety enforcement even as it defers hard choices on training data. In the U.S., the debate over �lawful use� is exposing how ambiguous contract language can ripple into reputational and market risk.
The uncomfortable truth from the Busch episode is that the tools are already in the wild. As Mercea notes, self-regulation tends to follow public outcry and regulatory threats; waiting for either is a costly strategy. Companies that pair clear guardrails with traceable data practices will be better positioned when the next manipulated image – or lawsuit – hits their notifications.