Imagem principal:Tara Winstead / Pexels
AI Scammers Outperform Humans in New Cybersecurity Study
New research suggests AI-powered fraudsters are more persuasive and scalable than human scammers, raising concerns about the future of digital deception. Cybersecurity Insiders reports that AI-driven scams may achieve higher success rates due to their ability to mimic human behavior with greater consistency and adaptability.
The claim that AI scammers may be more effective than their human counterparts has surfaced in a new study, prompting urgent questions about the evolving nature of digital fraud. While scams have long relied on human psychology, the integration of artificial intelligence introduces a new dimension: machines that can learn, adapt, and personalize interactions in real time. This development is not merely incremental—it represents a potential paradigm shift in how deception is scaled across digital platforms. To assess the validity and implications of this claim, this investigation synthesizes available reporting, examines the mechanisms behind AI-driven fraud, and evaluates expert responses to the growing threat.
—
AI Scammers Outperform Humans in Fraud Effectiveness Study
Cybersecurity Insiders reports that a recent study found AI-powered scammers to be more effective than human scammers in several key metrics, including response rates, message personalization, and persistence in conversation. The study, which has not yet been peer-reviewed, suggests that AI systems can maintain coherent, contextually appropriate dialogue over longer interactions, a capability that human scammers often struggle to sustain due to fatigue or limited cognitive bandwidth.
According to Cybersecurity Insiders, the research team used large language models fine-tuned for social engineering tasks and evaluated their performance against human scammers in controlled simulations. The AI systems were reported to generate more persuasive and emotionally resonant messages, adapt their tone based on user responses, and avoid common linguistic patterns that flag human scammers—such as grammatical errors or unnatural phrasing.
—
O que os Especialistas em Cibersegurança Relataram: Principais Descobertas e Metodologia
Cybersecurity Insiders describes a study in which AI-driven scammers were tested against human scammers across multiple fraud scenarios, including phishing, romance scams, and tech support fraud. The AI systems were trained on datasets of real human interactions and optimized for deception rather than transparency. The study reportedly found that AI scammers achieved higher success rates in eliciting sensitive information and maintaining user engagement over extended conversations.
The methodology involved simulating scam interactions in a controlled environment, with both AI and human agents attempting to extract personal or financial details from test subjects. Cybersecurity Insiders notes that the AI systems were particularly effective in scenarios requiring emotional manipulation, such as romance scams, where sustained, empathetic dialogue is critical. The study also highlighted the scalability of AI scams: once deployed, a single AI agent can conduct thousands of simultaneous conversations without fatigue, a feat impossible for human operators.
The report emphasizes that the AI models were not merely mimicking human speech but were engineered to exploit known cognitive biases—such as the tendency to trust consistent, coherent communication—while avoiding detectable red flags. Cybersecurity Insiders quotes a researcher involved in the study as saying, “The AI doesn’t just sound human; it sounds *more* human than most people do online.”
—
Cross-Referencing the Claim: Como as Fraudes com IA se Comparam às Fraudes Humanas
While Cybersecurity Insiders presents the most detailed account of the study’s findings, the broader cybersecurity community has long acknowledged the potential for AI to enhance fraudulent schemes. Earlier reporting from industry analysts at Gartner and Forrester has warned that AI could lower the barrier to entry for scammers by automating the most labor-intensive aspects of deception, such as crafting personalized messages or managing multiple conversations at once. However, Cybersecurity Insiders is the first to provide a structured comparison between AI and human scammers in a controlled setting.
For instance, while Gartner’s 2025 report on “AI in Social Engineering” highlighted the efficiency gains from automation, it did not quantify a measurable advantage in success rates over human scammers. In contrast, Cybersecurity Insiders claims that the study’s AI agents outperformed humans by a statistically significant margin in key engagement metrics. This discrepancy underscores the need for further peer-reviewed research to validate the findings and assess their generalizability across different fraud types and cultural contexts.
Another point of divergence is the role of detection. While Cybersecurity Insiders focuses on the AI’s ability to avoid detection through natural language fluency, other experts cited in industry forums argue that AI-generated content often leaves subtle digital fingerprints—such as unnatural pauses, repetitive phrasing, or metadata inconsistencies—that can be flagged by advanced detection tools. The study’s methodology, as described by Cybersecurity Insiders, does not appear to account for these detection vectors, suggesting that real-world effectiveness may depend heavily on the sophistication of both the scammer and the defender.
—
O Mecanismo por Trás da Eficácia dos Golpes com IA
Personalization at Scale
Cybersecurity Insiders reports that the AI scammers’ primary advantage lies in their ability to personalize interactions dynamically. Unlike human scammers, who may rely on pre-written scripts or limited improvisation, AI systems can analyze user inputs in real time and tailor responses to individual psychological profiles. For example, if a target mentions financial stress, the AI can adjust its messaging to emphasize urgency or opportunity, a tactic that Cybersecurity Insiders describes as “emotional micro-targeting.”
This capability is powered by large language models trained on vast datasets of human conversation, allowing the AI to mimic not just language patterns but also emotional cues. Cybersecurity Insiders notes that the AI’s responses were rated as more empathetic and coherent than those of human scammers in user surveys, a finding that aligns with broader research on AI’s ability to simulate emotional intelligence.
Persistence Without Fatigue
Another critical mechanism is endurance. Human scammers are limited by time, energy, and attention span, often abandoning targets after a few exchanges or making errors due to fatigue. Cybersecurity Insiders describes the AI systems as capable of maintaining conversations for hours or even days without degradation in quality, a feature that significantly increases the likelihood of success in prolonged scams such as romance fraud or investment pitches.
The study reportedly found that AI scammers were particularly effective in scenarios where trust is built gradually, such as in tech support scams or fake job offers. The AI’s ability to remember context across sessions—without the limitations of human memory—allows it to pick up conversations seamlessly, even after days of inactivity.
Adaptive Deception
Cybersecurity Insiders highlights the AI’s capacity for adaptive deception, where the system learns from each interaction and refines its approach. For example, if a target expresses skepticism about a particular claim, the AI can pivot to alternative narratives or introduce supporting “evidence” generated on the fly. This adaptability makes the scam harder to resist, as the AI can counter objections in real time rather than relying on a static script.
This mechanism mirrors findings from behavioral psychology, where adaptive persuasion has been shown to increase compliance rates. However, Cybersecurity Insiders does not provide specific data on how often the AI systems successfully adapted versus failed, leaving open questions about the limits of this capability in real-world conditions.
—
Quem É Mais Vulnerável a Esquemas de Fraude com IA?
Cybersecurity Insiders identifies several demographic groups as particularly susceptible to AI-driven scams, based on the study’s simulations and prior research on fraud victimization. Older adults, for instance, were reported to be more likely to fall for AI-generated romance scams due to higher levels of loneliness and lower familiarity with digital deception tactics. The study suggests that AI scammers’ ability to simulate long-term emotional investment—such as remembering personal details and expressing concern over time—exploits vulnerabilities common in this population.
Younger adults, while more digitally savvy, were found to be vulnerable to AI-powered phishing and investment scams, particularly when the AI mimicked trusted authorities such as banks or government agencies. Cybersecurity Insiders notes that the AI’s use of official-sounding language and plausible narratives made it harder for targets to distinguish between legitimate and fraudulent communications.
The study also highlights the role of social isolation as a risk factor. Individuals who spend significant time online—such as remote workers or gamers—were reported to be more likely to engage with AI scammers, as the AI’s persistent, personalized interactions can create a false sense of connection. Cybersecurity Insiders quotes a psychologist specializing in cybercrime as saying, “The AI doesn’t just mimic a person; it mimics the *idea* of a person who cares about you.”
—
How AI Scams Spread: Channels and Tactics Used by Fraudsters
Cybersecurity Insiders outlines several channels through which AI scams are reportedly being deployed, with social media platforms, messaging apps, and dating sites identified as primary vectors. The study describes AI scammers as using a mix of automated outreach and human-AI hybrid models, where an initial AI-driven conversation is later handed off to a human operator for final persuasion.
On social media, AI scammers are reported to use fake profiles that engage users in private messages, often posing as acquaintances, celebrities, or professionals offering lucrative opportunities. Cybersecurity Insiders notes that these profiles are designed to appear authentic, with AI-generated photos, bios, and posting histories that evolve over time to avoid detection by platform algorithms.
Apps de mensagens, especialmente aquelas com criptografia de ponta a ponta, são citadas como outro canal de alto risco. O estudo descreve golpistas de IA usando essas plataformas para conduzir conversas prolongadas e individuais, onde a capacidade da IA de manter o contexto e adaptar as mensagens é particularmente eficaz. Apps de namoro também são destacados como alvos principais, com sistemas de IA supostamente automatizando as fases iniciais de golpes românticos antes de passar para chamadas de voz ou vídeo para aprofundar a decepção.
Cybersecurity Insiders também adverte que as fraudes com IA estão cada vez mais integradas a ecossistemas mais amplos de cibercrime, nos quais dados roubados em uma fraude são usados para alimentar outra. Por exemplo, detalhes pessoais extraídos de uma fraude de phishing podem ser inseridos em um sistema de IA para criar mensagens de fraude romântica mais convincentes, criando um ciclo de retroalimentação que amplia o alcance e a eficácia da fraude.
—
Bandeiras Vermelhas e Lista de Desmistificação: Identificando Golpes Gerados por IA
While AI scammers are designed to avoid traditional red flags, Cybersecurity Insiders provides a checklist of warning signs that may indicate AI involvement. These include:
- Unnatural consistency: Messages that maintain perfect grammar, spelling, and tone across long conversations, without the natural variations seen in human speech.
- Overly personalized flattery: Excessive compliments or personal details that seem too precise or intrusive, especially in early interactions.
- Urgência sem contexto: Requests for immediate action—such as transferring money or sharing sensitive information—without a clear, verifiable reason.
- Refusal to engage in video calls: Persistent excuses for avoiding face-to-face communication, particularly in romance or job scams.
- Scripted responses: Answers that feel rehearsed or evasive when challenged, often repeating phrases or deflecting questions.
- Metadados inconsistentes: Inconsistencies in timestamps, geolocation, or device information that suggest automated rather than human interaction.
- Ofertas boas demais para ser verdade: Investment opportunities, prizes, or job offers that promise unusually high returns with minimal risk or effort.
Cybersecurity Insiders advises cross-referencing any suspicious communication with known legitimate sources—such as official websites or verified contact information—for the entity or individual in question. The report also recommends using reverse image searches to check the authenticity of profile pictures and searching for unique phrases from the message in search engines to identify potential scam templates.
—
Expert and Institutional Responses to Rising AI Fraud Threats
Cybersecurity Insiders includes reactions from cybersecurity professionals and law enforcement agencies, many of whom emphasize the urgency of addressing AI-driven fraud. A spokesperson for the FBI’s Internet Crime Complaint Center (IC3) is quoted as saying, “The use of AI in scams represents a significant escalation in the sophistication of cybercriminals. We are seeing a rapid increase in complaints involving AI-generated voice and text, and we urge the public to remain vigilant.”
The report also cites a statement from Europol’s European Cybercrime Centre (EC3), which warns that AI scams are likely to become more prevalent as the technology becomes more accessible. EC3 highlights the challenge of attributing AI-generated scams to specific actors, noting that the anonymity provided by AI tools complicates investigations and prosecutions.
Cybersecurity industry analysts, such as those at Mandiant and CrowdStrike, are reported to be developing new detection tools specifically designed to identify AI-generated content. These tools reportedly analyze linguistic patterns, metadata, and behavioral cues to flag potential scams. However, Cybersecurity Insiders notes that as detection methods improve, scammers are likely to adapt by incorporating more sophisticated AI models or hybrid human-AI approaches.
Regulatory bodies are also beginning to respond. The report mentions that the U.S. Federal Trade Commission (FTC) has issued warnings about AI scams and is exploring new rules to hold platforms accountable for hosting fraudulent AI-generated content. Meanwhile, the European Union’s AI Act, which includes provisions on transparency for AI-generated content, is cited as a potential model for mitigating the risks of AI-driven deception.
—
Original Analysis: Why AI Scams May Signal a New Era of Digital Deception
Taken together, the reporting on AI scammers suggests that we are witnessing the emergence of a new class of digital deception, one that combines the scalability of automation with the psychological nuance of human manipulation. Unlike traditional scams, which rely on volume and repetition, AI scams leverage adaptability and personalization to achieve higher success rates with fewer attempts. This shift is not merely quantitative—it is qualitative, altering the fundamental dynamics of trust and persuasion in digital spaces.
One of the most concerning implications is the democratization of fraud. As AI tools become more accessible, the barrier to entry for sophisticated scams is lowered. Cybercriminals no longer need to master the art of persuasion or invest significant time in grooming victims; instead, they can deploy AI systems that do the heavy lifting. This could lead to an explosion in the volume and variety of scams, overwhelming both individuals and institutions.
Another critical factor is the arms race between scammers and defenders. As detection tools improve, scammers are likely to adopt more advanced AI models, creating a feedback loop where each side pushes the other to greater sophistication. This dynamic could disproportionately affect vulnerable populations, who may lack the resources or awareness to keep pace with evolving tactics.
Finally, the rise of AI scams raises broader questions about the integrity of digital communication. If AI can convincingly mimic human interaction, how do we distinguish between authentic and synthetic personas? The implications extend beyond fraud, touching on issues of misinformation, impersonation, and the erosion of trust in online spaces. Addressing these challenges will require a multi-stakeholder approach, involving technology companies, policymakers, and civil society.
—
Actionable Steps: Protecting Yourself from AI-Powered Scams
Cybersecurity Insiders offers several recommendations for individuals and organizations to reduce their risk of falling victim to AI scams:
- Verify before you trust: Always independently verify the identity of anyone contacting you online, especially if they ask for money or sensitive information. Use official contact details from trusted sources rather than those provided by the potential scammer.
- Limit exposure: Be cautious about sharing personal details on social media and dating platforms. The more information available about you, the easier it is for scammers—AI or otherwise—to craft convincing messages.
- Use autenticação multifator (MFA): Enable MFA on all accounts to add an extra layer of security against unauthorized access, even if scammers obtain your login credentials.
- Eduque-se e aos outros: Stay informed about the latest scam tactics and share this knowledge with friends and family, particularly older adults who may be more vulnerable.
- Relate atividade suspeita: If you encounter a potential AI scam, report it to the relevant platform and to authorities such as the FBI’s IC3 or your local cybercrime unit. This helps build a database of tactics and can aid in investigations.
- Use detection tools: Consider using browser extensions or apps that flag potential scam websites or messages. While no tool is foolproof, they can provide an additional layer of defense.
- Be skeptical of urgency: Scammers often create a sense of urgency to pressure victims into acting quickly. Take a step back and assess the situation before responding.
For organizations, Cybersecurity Insiders recommends implementing advanced email and message filtering systems that can detect AI-generated content, as well as conducting regular training for employees on recognizing and reporting scams. The report also advises businesses to review their customer verification processes, particularly for high-risk transactions.
—
FAQ: AI Scammers, Effectiveness, and Fraud Prevention
What makes AI scammers more effective than humans?
AI scammers leverage large language models to generate highly personalized, coherent, and emotionally resonant messages at scale. Unlike humans, they can maintain consistent interactions over long periods without fatigue, adapt their tactics in real time, and avoid common linguistic red flags that expose human scammers.
Are AI scams already happening in the real world?
Yes. Cybersecurity Insiders reports that AI-driven scams are being deployed across social media, messaging apps, and dating platforms. While the study focuses on simulations, there is evidence that scammers are already using AI tools to enhance their operations, particularly in romance and phishing scams.
Can AI-generated scams be detected?
Detection is possible but increasingly challenging as AI models improve. Cybersecurity Insiders highlights several red flags, such as unnatural consistency in messaging and refusal to engage in video calls. However, scammers are likely to adapt by incorporating more sophisticated AI or hybrid models, making detection an ongoing arms race.
Who is most at risk from AI scams?
The study identifies older adults, socially isolated individuals, and those who spend significant time online as particularly vulnerable. Older adults may be more susceptible to romance scams due to loneliness, while younger adults may fall for investment or phishing scams that mimic trusted authorities.
O que posso fazer para me proteger de golpes com IA?
Cybersecurity Insiders recomenda verificar identidades de forma independente, limitar a exposição de dados pessoais, ativar a autenticação multifator, manter-se informado sobre táticas de fraude e utilizar ferramentas de detecção. Para organizações, sistemas avançados de filtragem e treinamento de funcionários são etapas críticas.
—