This is a research prototype. The data and analyses are preliminary and not yet validated — we'd welcome your .
Threat actors have leveraged AI-generated content, including deepfake videos and imagery, to scale large-scale investment and tech support scams. By integrating these AI assets with Keitaro traffic distribution systems, attackers successfully cloaked malicious landing pages from security scanners while targeting specific demographics with fraudulent investment opportunities. The campaign spanned over 15,500 domains, impersonating legitimate news outlets and public figures to deceive victims into providing financial information.
Researchers reported that threat actors abused Keitaro Tracker infrastructure to cloak and route AI-themed investment scams across about 15,500 domains. The campaigns allegedly used AI-trading claims, deepfake videos or imagery, generative AI-produced copy and visuals, spoofed news pages, and fraudulent trading platforms to target potential victims while hiding scam content from scanners and reviewers.
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 highly realistic fake news articles and deepfake broadcasts to mislead users into believing the scams were legitimate.
The incident involves threat actors using generative AI deepfakes, images, and text to execute large-scale financial fraud and investment scams targeting individual victims.
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 risk was caused by human threat actors intentionally using deployed generative AI tools to create deceptive assets for financial scams.
Added categories peak at 4: Severe (Epistemic Harm, Inferred)
Threat actors leveraged generative AI and deepfakes of public figures, including the Canadian Prime Minister, to scale financial fraud across 15,500 domains. While the campaign demonstrates a highly sophisticated and evolved cybercrime capability, its impact is classified as minor from a national security perspective as it does not target critical infrastructure or systemic state functions.
The AI systems (such as OpenAI's DALL-E and deepfake video generators) operated as tools, performing specific, well-defined functions like generating images or synthesizing video based on direct human inputs without autonomous planning or context retention.
Risk Level 3: Limited Risk. The report describes the use of 'AI-generated content such as deepfakes' and 'chatbots' (specifically, deepfake video ads impersonating news anchors and fake chatbots used as landing pages), which fall under the transparency obligations of Limited Risk systems.
The primary use of AI described in the reports is the generation of deepfake videos of news anchors and public figures to deceive victims, alongside the generation of deceptive images and marketing copy.
The reports provide detailed, empirical cybersecurity research from Infoblox and Confiant, tracking over 15,500 domains and documenting specific threat actor groups and their TTPs. The role of AI in generating deepfakes, copy, and images is clearly documented, and the connection to the Keitaro TDS is thoroughly analyzed.
The incident could be viewed primarily as a traditional cybercrime/phishing campaign where AI is merely an incidental tool, rather than an inherent failure of an AI system itself.
Technology
Generative AI Lure Generation
Generative AI used to mass-produce convincing scam headlines and copy.
Deepfake Video Generation
AI-generated deepfakes impersonate anchors to build false trust.
AI Text-to-Image Generators
DALL-E used to create realistic assets for tech support scam lures.
Data Inputs
Deepfake Source Material
Using real video and audio of public figures to generate deepfakes.
AI-Themed Marketing Lures
Promoting fake AI trading algorithms as data inputs to lure victims.
Impersonated Brand Assets
Using popular brand logos and names as inputs for AI-generated ads.
Human Factors
Overtrust in AI Technology
Victims trust fake claims of highly profitable AI-driven trading.
Susceptibility to Deepfakes
Users find highly realistic deepfake video ads convincing and urgent.
Susceptibility to Chatbots
Victims trust low-friction fake AI chatbots that qualify them for scams.
Process and Methods
Evasion via Domain Cloaking
Using Keitaro TDS to hide AI scam content from security reviewers.
Scalable AI Lure Production
Programmatic AI asset generation allows rapid scaling of scam pages.
Targeted Ad Distribution
Using Bigo and Facebook Ads to deliver AI deepfakes to target audiences.
Management
Inadequate Ad Platform Review
Ad networks fail to detect AI scams due to weak automated review tools.
Abuse of Commercial Trackers
Keitaro tracker licensed to threat actors or run via illicit copies.
Slow Abuse Mitigation
Security tools struggle to keep up with the scale of AI domain rotation.
Regulatory Environment
Lack of Deepfake Regulation
Absence of strict controls on creating and distributing deepfakes.
Weak Ad Network Verification
Ad platforms fail to verify the authenticity of AI-generated ads.
AI system involved
Identified by automated extraction; can differ from the alleged developers above.
