This is a research prototype. The data and analyses are preliminary and not yet validated — we'd welcome your .
Hack Club founder PERSON_001 was targeted by a sophisticated AI-driven phishing attack where scammers impersonated Google support staff. The attackers used AI voice cloning to sound like real engineers and exploited Google's own infrastructure, including the Google Assistant calling system and the g.co subdomain, to gain the victim's trust. The victim narrowly avoided account compromise by identifying inconsistencies in the scammers' story. Google has since suspended the account used in the scam and is working to harden its defenses against such abuse.
Hack Club founder Zach Latta was reportedly targeted in an AI-assisted phishing attack exploiting Google’s own systems. Cybercriminals allegedly spoofed a legitimate Google Assistant number and used voice cloning to impersonate support staff, claiming unusual login activity from Frankfurt. They reinforced the deception by sending a real email from Google’s Workspace domain. The attack, which relied on Google’s g.co subdomain, was thwarted when inconsistencies raised Latta’s suspicions.
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 attackers exploited vulnerabilities in Google's infrastructure, including spoofing Google Assistant caller ID and abusing the g.co subdomain registration process.
The scammers used AI-powered voice cloning to impersonate Google support staff in a highly targeted attempt to defraud specific victims and steal their account credentials.
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 initiated by human cybercriminals who intentionally used AI voice generation and infrastructure exploits to conduct a phishing scam.
Cybercriminals leveraged sophisticated AI voice cloning and Google infrastructure exploits to conduct highly convincing phishing scams against Gmail users. While resulting in individual financial losses, the incident represents an evolved cybercrime capability rather than a direct threat to national security, highlighting the growing challenge of AI-enabled social engineering.
The AI system involved is Google Assistant, which acts to augment human capabilities and requires human oversight, though in this case, scammers also used AI voice generation tools to mimic human support staff.
Limited Risk: The incident involves AI-generated content such as deepfakes, specifically AI-powered phone calls mimicking human voices to deceive users.
The primary AI capability leveraged by the scammers was highly realistic voice generation to mimic human support engineers. Google Assistant was also abused as part of the calling infrastructure.
The reports provide consistent, detailed accounts of the phishing methodology, including specific technical exploits like the g.co subdomain abuse and Google Assistant caller ID spoofing. Multiple independent sources and victims confirm the pattern of the attack, leaving little ambiguity about the nature of the threat.
The incident could be viewed purely as a traditional social engineering scam, but the integration of highly realistic AI voice cloning and infrastructure exploits makes it a distinct AI-enabled threat.
Technology
Realistic AI Voice Generation
AI-powered voices mimicking humans made the phishing call highly realistic.
Caller ID Spoofing
Attackers spoofed Google Assistant number to display a Google caller ID.
Abuse of g.co Subdomain
Attackers used g.co subdomain to send authentic-looking emails.
Data Inputs
Unverified Sign-up Data
Google allowed g.co references at sign-up without domain verification.
Genuine 2FA Code Injection
Scammers triggered real 2FA codes to build trust during the call.
Human Factors
Reliance on Traditional Rules
Victims trusted verified phone numbers and domains to confirm legitimacy.
Cognitive Overload
Conflicting information and real 2FA codes confused the victim.
Process and Methods
Lack of Domain Verification
No process existed to verify ownership of g.co subdomains at sign-up.
Inadequate Voice Authentication
No mechanism was in place to verify if support calls were from humans.
Management
Delayed Defense Hardening
Google only hardened defenses after the sophisticated scam was reported.
Poor Abuse Risk Assessment
Management overlooked the risk of g.co subdomain abuse for phishing.
AI system involved
Identified by automated extraction; can differ from the alleged developers above.
