Nine Network's AI Alters Lawmaker Georgie Purcell's Image Inappropriately

The Nine Network used Photoshop's Generative Expand AI tool to resize an image of lawmaker Georgie Purcell, inadvertently altering her attire to appear more revealing. This error, claimed to result from the AI's automation, led to public criticism and an apology from the network.

Risk classification

• Primary risk domain: 1 Discrimination & Toxicity
• Primary risk subdomain: 1.1 Unfair discrimination and misrepresentation

The AI system generated an output that unfairly represented and sexualized a female politician, reinforcing gender-based stereotypes and misrepresenting her actual appearance.

Additional risk subdomains

• 1.2 Exposure to toxic content: The incident exposed the MP and the viewing public to inappropriate, sexualized synthetic content of a public figure.
• 7.3 Lack of capability or robustness: The AI tool failed to perform robustly and reliably, generating an inaccurate and highly inappropriate image during a routine resizing task.

Causal factors

• Entity: AI
• Intent: Unintentional
• Timing: Post-deployment

The incident was caused by the automated output of Adobe Photoshop's Generative Expand tool, which unintentionally generated a sexualized depiction of the MP during a resizing process.

EU AI Act risk tier

• Risk tier: 3 Limited Risk

Limited Risk: The incident involves the generation and broadcast of altered synthetic media (a deepfake/manipulated image of a person), which falls under transparency obligations for AI-generated content.

AI system and alleged parties

• AI system: Photoshop Generative Expand (Adobe)
• AI purpose: Image Retouching; Image Generation
• Behaviour type: Assistant
• Alleged developer: Adobe
• Alleged deployer: Nine Network
• Alleged harmed parties: Georgie Purcell

Harm severity

Severity ratings are based on this scale.

Highest direct severity in any category: Minor. Severity is scored from Negligible to Catastrophic in each harm category, for harm the reports describe as caused directly or indirectly by the AI system.

• Physical: direct Negligible, indirect Negligible
• Infrastructure: direct Negligible, indirect Negligible
• Property: direct Negligible, indirect Negligible
• Financial: direct Negligible, indirect Negligible
• Environmental: direct Negligible, indirect Negligible
• Malicious content: direct Minor, indirect Minor
• Differential treatment: direct Minor, indirect Minor
• Civil rights: direct Negligible, indirect Minor
• Democracy: direct Negligible, indirect Negligible
• Privacy: direct Negligible, indirect Negligible
• Psychological: direct Minor, indirect Negligible
• Epistemic: direct Minor, indirect Minor
• Child sexual exploitation and abuse: direct Negligible, indirect Negligible

Malicious content

Reported: The report explicitly describes toxic content created and spread directly and indirectly by the incident.

Directly caused: The AI tool directly generated a sexualized and objectified image of Georgie Purcell, which was broadcast to the public on a nightly news bulletin.

Indirectly caused: The broadcast of the altered image was subsequently weaponized by online abusers and shooters to send 'hateful and violent messages' telling her to 'get back on the pole'.

Differential treatment

Reported: The report explicitly describes differential treatment of individuals based on gender caused directly and indirectly by the incident.

Directly caused: The AI tool generated a highly sexualized and objectified image of a female politician, a form of misrepresentation that Purcell and other commentators noted would never happen to male colleagues.

Indirectly caused: The incident highlights systemic sexism and the disproportionate targeting of women in public life with sexualized synthetic media.

Civil rights

Reported: The report explicitly describes violations of human and civil rights caused directly and indirectly by the incident.

Directly caused: The non-consensual alteration and sexualization of Purcell's body violated her personal dignity, bodily integrity, and moral rights.

Indirectly caused: The incident contributes to a hostile environment for women in politics, potentially discouraging young women from pursuing public office.

Psychological

Reported: The report explicitly describes psychological harm to the affected individual, Georgie Purcell, caused directly by the incident.

Directly caused: Georgie Purcell reported feeling 'really, really uncomfortable' and 'violated' upon seeing her body altered on television, which triggered her PTSD from past image-based abuse.

Epistemic

Reported: The report explicitly describes epistemic harm caused directly and indirectly by the incident.

Directly caused: The news broadcast presented a fabricated, digitally altered image of a public official as a real photograph, misleading viewers about her actual appearance.

Indirectly caused: The incident contributes to growing public skepticism about the credibility of journalism and sows doubt about the authenticity of media images in general.

People affected

• Occurrences reported: 1
• People reportedly harmed: 1
• People reportedly exposed: 1

Potential causes

Management

• Prioritizing Speed over Safety: Management focused on operational efficiency over editorial accuracy.
• Avoidance of Accountability: Network leaders blamed the AI software instead of taking responsibility.

Technology

• AI Bias in Generative Expand: Photoshop's AI tool tends to generate sexualized outfits for women.
• Flawed Generative Expansion: The tool made incorrect guesses when expanding a cropped image.

Data Inputs

• Biased Training Datasets: Adobe Firefly was trained on stock images that reinforce gender stereotypes.
• Cropped Source Photo Input: Inputting a tightly cropped image forced the AI to guess missing details.

Human Factors

• Lack of Operator Verification: The designer failed to review or correct the AI's sexualized output.
• Overreliance on AI Automation: Staff trusted the automated tool without checking the final output.

Process and Methods

• Inadequate Editorial Controls: No process was in place to verify the accuracy of AI-modified images.
• Undisclosed Use of AI Tools: Nine News used Generative Expand without declaring it or flagging the edits.

Regulatory Environment

• Lack of Mandatory AI Safeguards: Weak regulatory frameworks in Australia let AI risks go unmanaged.
• Absence of Watermarking Codes: Lack of required watermarking or labeling for AI-altered media.

Information quality

• Classification confidence: High
• Reason for confidence: The reports provide comprehensive, multi-perspective coverage of the incident, including statements from the victim, the news network, the software developer (Adobe), and independent technical experts. This allows for a highly reliable assessment of the events, the technology used, and the resulting harms.
• Ambiguities identified: There is a dispute between Nine News (claiming the alteration was fully automated during resizing) and Adobe (stating that human intervention and approval were required to make these specific changes).
• Alternative interpretations: One could interpret the incident as purely human-driven graphic design manipulation rather than an AI failure, though the technical consensus points to the use of Photoshop's generative AI tool.

An Australian news network broadcast a digitally altered, sexualized image of state MP Georgie Purcell generated via Photoshop's Generative Expand tool. While raising significant ethical concerns about generative AI in journalism and gender-based harassment, the incident remains a localized media error with negligible implications for national security.

• Overall national security impact: Minor
• Scope: Single nation
• Primary target: Australia
• Alleged perpetrator: Nine News Melbourne

Threat characteristics

• Imminence: Long-term. The immediate broadcast incident is resolved, but the strategic concern regarding generative AI in media remains an ongoing issue.
• Autonomy: Human-supervised. The AI tool acted as an assistant to a human operator, requiring human initiation and final broadcast approval, though automation led to the error.
• Novelty: Evolved capability. Represents a significant advancement in the ease and speed of image manipulation using generative AI compared to traditional manual editing techniques.

Impact by dimension

• Physical security: Negligible. No physical threat to critical infrastructure, systems, or human safety was reported in this incident.
• Information security: Negligible. The incident was a localized media broadcast error and did not involve a coordinated foreign information warfare campaign or intelligence compromise.
• Sovereignty: Negligible. The incident did not disrupt core government operations, electoral systems, or state sovereignty, despite affecting an individual Member of Parliament.
• Economic security: Negligible. No threats to strategic industries, financial systems, or national technological advantages were present.
• Societal stability: Minor. The incident caused direct psychological harm to an MP and highlighted issues of gender-based harassment, but did not threaten large-scale social cohesion or civil liberties.