This is a research prototype. The data and analyses are preliminary and not yet validated — we'd welcome your .
A Russian-linked scam network used AI-generated deepfakes of journalist Maria Ressa to promote cryptocurrency scams and spread disinformation on Facebook and Microsoft Bing. The operation involved creating impostor websites mimicking Rappler and CNN Philippines to lend credibility to the fraud. While platforms removed the specific content, the report highlights the ongoing challenge of 'whack-a-mole' in mitigating AI-enabled disinformation and financial fraud.
A purportedly Russian-linked scam network allegedly circulated an AI-generated deepfake of journalist Maria Ressa on Facebook and Microsoft Bing. The video reportedly manipulated a 2022 interview to falsely depict Ressa endorsing cryptocurrency. Fraudulent websites allegedly impersonating Rappler and CNN Philippines reportedly amplified the scam. Investigators reportedly linked the operation to a fraudulent network targeting Filipino audiences. Meta and Microsoft removed the content, but similar scams allegedly continue to resurface.
Domain classification, causal taxonomy, severity scores, and national security assessments were LLM-classified and may contain errors.
Using AI systems to gain a personal advantage over others such as through cheating, fraud, scams, blackmail or targeted manipulation of beliefs or behavior. Examples include AI-facilitated plagiarism for research or education, impersonating a trusted or fake individual for illegitimate financial benefit, or creating humiliating or sexual imagery.
Also classified as
The campaign generated false news articles and videos to mislead the public about Maria Ressa's involvement in a cryptocurrency scam.
The incident involved using AI-generated voice clones and deepfakes to impersonate a trusted individual for financial benefit through a cryptocurrency scam.
Human
Due to a decision or action made by humans
Intentional
Due to an expected outcome from pursuing a goal
Post-deployment
Occurring after the AI model has been trained and deployed
The incident was caused by a Russian scam network intentionally deploying AI-generated deepfakes to defraud users.
Added categories peak at 3: Substantial (Epistemic Harm, Indirect)
A Russian-linked scam network utilized AI-generated deepfakes and voice clones of Nobel laureate Maria Ressa to run cryptocurrency fraud campaigns targeting Philippine audiences. While demonstrating evolved capabilities in synthetic media generation and ad-network evasion, the incident's impact remains confined to localized financial fraud and reputational damage, posing minor national security concerns.
The AI system was used as a tool to perform a specific, well-defined function of generating a cloned voice and syncing it to a video based on direct inputs, without autonomous planning.
Limited Risk: The report describes the creation of AI-generated deepfakes, which are subject to transparency obligations under the EU AI Act to ensure users are aware they are interacting with synthetic content.
The primary function of the AI system was to generate a deepfake video of Maria Ressa, utilizing voice generation to clone her voice.
The report is highly detailed, coming directly from Rappler (the targeted organization) and supported by a digital forensic investigation by Qurium, providing clear evidence of the scam's mechanics and impact.
None
Technology
Open-Source AI Misuse
Open-source LLMs allow bad actors to modify code and strip watermarks.
Watermark Removal Tech
Existing tools can easily remove watermarks designed to label AI content.
Realistic AI Voice Generation
Generative AI voice cloning mimics authoritative voices to deceive users.
Data Inputs
Scraped Media Data
Scraped original video and audio utilized to train and generate deepfakes.
Lack of Provenance Data
Absence of secure metadata on original media allowed unauthorized manipulation.
Human Factors
User Inattention
Users not paying attention fail to notice lip-syncing errors in deepfakes.
Trust in Authority Figures
Scammers exploit trust in journalists' reputations to make scams believable.
Process and Methods
Weak Ad Review Processes
Bing and Facebook allowed fraudulent ads promoting deepfakes to be published.
Slow Platform Response
Delayed takedowns allowed deepfakes to garner tens of thousands of views.
Whack-a-Mole Domain Tactics
Scammers rapidly register new domains to bypass platform domain bans.
Management
Inadequate Ad Monitoring
Platforms failed to proactively monitor and block AI-generated scam ads.
Profit from Disinformation
Platforms profit from ad revenues, reducing incentive for strict pre-screening.
Regulatory Environment
Lack of Platform Accountability
No strong legal frameworks to hold platforms liable for hosting fake ads.
Anonymity in Domain Registration
Scammers exploit loose domain registration rules to remain anonymous.
AI system involved
An AI system was involved, but the reports do not identify it.
Identified by automated extraction; can differ from the alleged developers above.
