Purportedly AI-Powered 007TG Platform Reportedly Used by Scam Networks Linked to $75 Million in Proceeds

007TG was a software suite reportedly used by scam compounds to run fraud across social media, using AI-assisted messaging and translation to help operators manage targets while concealing their identities. Wallet records linked the platform to at least four known scam networks that collectively received at least $75 million, though the evidence does not show how much was directly attributable to 007TG.

Risk classification

• Primary risk domain: 4 Malicious actors
• Primary risk subdomain: 4.3 Fraud, scams, and targeted manipulation

The primary aspect of the incident is the coordinated use of AI systems to conduct targeted romance scams and financial fraud for personal and organizational gain.

Additional risk subdomains

• 5.1 Overreliance and unsafe use: Victims anthropomorphized and placed trust in the AI-generated personas, leading to severe emotional and material dependence.

Causal factors

• Entity: Human
• Intent: Intentional
• Timing: Post-deployment

The harm is caused by human malicious actors intentionally abusing deployed AI models to conduct large-scale financial fraud and monitor trafficked workers.

EU AI Act risk tier

• Risk tier: 1 Unacceptable

Unacceptable Risk: The report describes 'AI used for subliminal manipulation or behavioral influence leading to harm' and 'exploitative systems targeting vulnerabilities' where scammers use AI to 'make each victim fall in love' and 'deceive targets' through 'manipulative language'.

AI system and alleged parties

• AI system: 007TG, ChatGPT, Gemini, KT Smart Translation (Google, OpenAI)
• AI purpose: Chatbot; Translation
• Behaviour type: Assistant
• Alleged developer: OpenAI, Large language model developers, HiSeven Pte. Ltd., Google DeepMind, Google, Chatbot developers, AI-assisted scam platform developers, 007TG (Global Social Traffic Navigation)
• Alleged deployer: Transnational fraud rings, Scammers, Pig-butchering scammers, Cybercriminals
• Alleged harmed parties: Trafficking victims coerced into scam operations, Targets of scams, Privacy, Human trafficking victims, Global victims targeted by scams, Financial scam victims, Epistemic integrity

Harm severity

Severity ratings are based on this scale.

Highest direct severity in any category: Substantial. 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 Substantial
• Infrastructure: direct Negligible, indirect Negligible
• Property: direct Negligible, indirect Negligible
• Financial: direct Substantial, indirect Minor
• Environmental: direct Negligible, indirect Negligible
• Malicious content: direct Minor, indirect Negligible
• Differential treatment: direct Negligible, indirect Negligible
• Civil rights: direct Negligible, indirect Substantial
• Democracy: direct Negligible, indirect Negligible
• Privacy: direct Minor, indirect Negligible
• Psychological: direct Minor, indirect Minor
• Epistemic: direct Minor, indirect Negligible
• Child sexual exploitation and abuse: direct Negligible, indirect Negligible

Physical

Reported: The report explicitly describes severe physical harm to trafficked workers in the scam compounds.

Indirectly caused: Trafficked workers who failed to meet performance quotas tracked by AI monitoring tools were severely beaten and tortured. One worker, Ebisa, was permanently blinded in one eye, and another, Obinna Okeadu, died following a severe beating.

Inferred additional harm: It is highly likely that many of the other thousands of trafficked workers in these compounds suffered similar physical abuse for failing to meet AI-monitored quotas.

Financial

Reported: The report describes massive financial losses, including individual losses of $400,000 and scam networks raking in at least $75 million.

Directly caused: The average financial loss per occurrence is estimated at $75,000,000, representing the total documented illicit profits generated by the scam networks using these AI tools.

Indirectly caused: Individual victims like Colocousis lost $400,000, and trafficked workers had to pay ransoms of approximately $5,300 for their freedom.

Inferred additional harm: The total global financial losses from these AI-enabled scams likely run into hundreds of millions or billions of dollars.

Malicious content

Reported: The report explicitly describes the creation and spread of malicious, deceptive content used to execute scams.

Directly caused: AI models directly generated automated, deceptive replies and roleplay scripts to manipulate and defraud targets.

Inferred additional harm: Millions of automated, deceptive messages were likely generated and distributed globally across various social media platforms.

Civil rights

Reported: The report explicitly describes severe violations of human and civil rights, including human trafficking and forced labor.

Indirectly caused: Trafficked workers were held against their will, forced to commit crimes, and subjected to physical violence, with their performance and behavior monitored by AI tools.

Inferred additional harm: Thousands of other trafficked workers in Southeast Asian scam compounds likely have had their basic human rights violated under similar AI-monitored conditions.

Privacy

Reported: The report explicitly describes privacy violations through the use of AI-powered monitoring tools.

Directly caused: 007TG's CloudSeven used AI models to perform sentiment analysis and track emotional shifts in target responses, as well as monitoring employee behavior.

Inferred additional harm: Massive unauthorized collection and analysis of personal communication data from both victims and forced workers.

Psychological

Reported: The report explicitly describes severe psychological distress and trauma experienced by both scam victims and trafficked workers.

Directly caused: Scam victims suffered severe emotional manipulation and distress from interacting with highly convincing AI-generated personas.

Indirectly caused: Trafficked workers suffered extreme fear, anxiety, and trauma due to constant surveillance and the threat of physical violence.

Inferred additional harm: Thousands of other scam victims likely suffered severe depression, anxiety, and trauma from losing their life savings and being emotionally exploited.

Epistemic

Reported: The report explicitly describes epistemic harm through the generation of highly convincing fake personas and deceptive information.

Directly caused: GPT-4-powered roleplay chatbots and automated translation tools were used to fabricate convincing characters and technical roles to mislead victims.

Inferred additional harm: Widespread erosion of trust in online communications due to the proliferation of highly convincing, AI-generated fake personas.

People affected

• Occurrences reported: 1
• People reportedly harmed: 50000
• People reportedly exposed: 50000

Potential causes

Management

• Prioritizing Growth Over Safety: Tech companies lack business incentives to invest heavily in scam prevention.
• Weak Enforcement of Terms of Service: Failure to vigorously enforce terms of service banning fraudulent activities.
• Inadequate API Access Controls: Management failed to restrict API access for gray-market scam software.

Technology

• Abuse of LLMs for Automated Replies: Scammers use ChatGPT and Gemini to automate deceptive messaging at scale.
• AI Roleplay Chatbots for Deception: GPT-4 powered chatbots help scammers build highly convincing personas.
• AI-Powered Employee Monitoring: 007TG uses AI sentiment analysis to track victim emotional shifts and quotas.

Data Inputs

• Difficulty Distinguishing Scams: Scam prompts mimic legitimate uses like translation and basic writing.
• Evasion via Laundered IP Addresses: Scammers route traffic through US cloud services to hide physical locations.
• Lack of Real-Time Prompt Flags: Input data lacks real-time flags to detect manipulative romance scam scripts.

Human Factors

• Coerced Labor in Scam Compounds: Trafficked workers are physically forced to operate AI-powered scam systems.
• Target Vulnerability to AI Personas: Targets are deceived by highly realistic AI-generated personas and text.
• Cognitive Enablement of Scammers: AI tools enable low-skill scammers to assume highly technical roles.

Process and Methods

• Reactive Abuse Mitigation Policies: AI companies and ISPs mostly take action after abuse reports are received.
• Opaque Gray Market Software Sales: Software like KT and 007TG is sold via Telegram and paid with crypto.
• Inadequate Proactive Monitoring: Lack of active systems to monitor and block API misuse in real-time.

Regulatory Environment

• Lack of Legal and Business Incentives: US tech companies face no financial cost for facilitating cyber scams.
• Voluntary Cooperation Model in US: US regulators rely on voluntary cooperation rather than strict penalties.
• No Utility-Level Safety Mandates: Tech and telecom providers are not legally responsible for content safety.

Information quality

• Classification confidence: High
• Reason for confidence: The reports from AP and C4ADS provide extensive, highly detailed, and well-documented evidence of the AI models used, the specific software suites (KT and 007TG), the mechanisms of fraud, and the severe physical and financial harms caused. The findings are backed by blockchain analysis, device data, and first-hand interviews with victims and former scammers.
• Ambiguities identified: The exact proportion of scam operations fully automated by AI versus those requiring human copy-pasting is not fully quantified.
• Alternative interpretations: None. The role of AI in scaling and facilitating these scams is clear and undisputed in the reports.

Criminal syndicates in Southeast Asia are leveraging commercial LLMs to scale global financial fraud and monitor trafficked workers. While not a direct state-level threat, the integration of AI into transnational organized crime represents a substantial escalation in economic harm and human rights abuses.

• Overall national security impact: Substantial
• Scope: Multiple nations
• Primary target: No clear primary
• Other affected: United States, Myanmar, Philippines, and multiple other nations
• Alleged perpetrator: Southeast Asian criminal syndicates

Threat characteristics

• Imminence: Long-term. Represents an ongoing, systemic criminal capability and strategic law enforcement challenge rather than an immediate, acute crisis.
• Autonomy: Human-supervised. AI systems automate messaging and monitoring, but human operators remain in the loop to manage interactions and enforce quotas.
• Novelty: Evolved capability. Significant advancement of traditional cyber-enabled scamming and trafficking operations through the integration of commercial LLMs.

Impact by dimension

• Physical security: Minor. Physical abuse and torture of trafficked workers occurred within the compounds, indirectly driven by AI-powered performance and quota-tracking systems.
• Information security: Minor. Large-scale automated deception and persona-generation used for financial scams, though lacking state-sponsored political or intelligence-targeting objectives.
• Sovereignty: Negligible. No direct threat to state authority, territorial control, or core government decision-making processes.
• Economic security: Substantial. Industrialized AI-enabled fraud generating tens of millions in illicit profits globally, utilizing US-based tech and satellite infrastructure.
• Societal stability: Substantial. Severe human rights violations, including human trafficking and forced labor, systematically monitored and optimized using AI sentiment and behavior tracking.