The IA is no longer optional: the black market for access threatens companies

Artificial intelligence tools ceased to be an experiment to become a daily piece of productivity: from writing emails and generating code to summarizing research and automating business processes. Platforms such as ChatGPT, Claude, Microsoft Copilot or Perplexity are already a daily part of people and organizations, and in many cases have even been integrated into critical operating flows. When a technology becomes indispensable, its value grows - and not just for those who use it legitimately.

A recent analysis of the Flare threat intelligence company shows that, behind the comfort of these platforms, a clandestine market has appeared: premium accounts, API keys and access to payment services are announced and resold in closed forums and groups, especially on Telegram channels and Russian-speaking markets. Flare researchers have collected numerous ads and discussions that show how access to advanced IA services is packaged and offered as a product more within the digital criminal ecosystem. The Flare page is available for more context on your work: Flare, and those who wish to do so can explore the findings through the test options the company offers: free Flare test.

The ads are not anecdotal or isolated. In many cases, top-level subscriptions such as ChatGPT Plus or Pro, Anthropic Claude Pro, Perplexity Pro and packages including Microsoft Copilot with Office 365 accounts are sold. There are also offers that promise "unlimited access" or "complete API," claims that seek to attract buyers who wish to evade official restrictions. In addition, these accesses are often marketed along with other digital assets such as old mail accounts, virtual servers or RDP accesses, which facilitates their use in more complex fraudulent operations.

How do criminals get these accounts? Flare does not document each technique directly, but sales patterns point to several ways: keys and secrets exposed in public repositories or containers, committed credentials reused to take account control, massive creation of profiles with virtual numbers to skip verifications, abuse of promotional codes or test periods and even resale of API keys obtained by various means. Previous investigations have alerted about the escape of credentials and secrets in public containers, and good security practices on development platforms show how an exposed key can give access to sensitive services; seeing APis security and secret management documentation can help you understand the problem, for example in OpenAI security guides: OpenAI - Security.

The appeal to buy access in the black market is essentially economic and practical. A premium IA subscription can cost tens of dollars a month and, when multiple accounts are needed to automate fraud, test or evade controls, buying already provided access is easier and cheaper than recording and checking them one by one. In regions affected by sanctions or payment constraints - such as Russia, Iran or North Korea - these offers also serve as a shortcut to remove access barriers, offering ready-to-use accounts. In short, the resale of access reduces friction and expands the potential audience of those who can abuse technology.

Once in the hands of malicious actors, the capabilities of these platforms multiply the impact of their operations. Generative tools facilitate the mass writing of highly credible phishing emails, scam scripts, custom content for social engineering and even synthetic multimedia materials that allow more sophisticated suplantations. Europol already warns in its threat assessment about the increasing use of IA to automate and refine fraud and phishing campaigns: EU SOCTA 2025. Similarly, technical reports such as those of Palo Alto Networks describe how the attack chains incorporate content generated by IA to achieve more accuracy in victim manipulation: Unit 42 - Report. Even academic developments and publications from IA companies, such as Anthropic's reports on the detection and mitigation of improper uses, show that the sector is aware of the problem: Anthropic - Abuse report.

This commercial flow not only feeds scams; it also facilitates technical operations. With access to language models and code generation tools, actors with little experience can automate complex tasks, create attack infrastructure or generate exploits and scripts more quickly. In addition, the possibility of acquiring fragmented or "shared" access - where an account is used by multiple buyers - complicates traceability and increases the resilience of fraudulent campaigns against point blocks.

In front of this scenario, organizations are not defenceless. There are concrete and practical measures that significantly reduce the risk that their IA accounts will be compromised or appear in illicit markets. These include forcing multi-factor authentication in all accounts used for business projects, choosing business plans and environments that provide administrative controls and business records, and strictly managing the keys and secrets that connect APIs applications. It is also essential to monitor unusual access patterns and behaviour in services, to rotate credentials regularly and to restrict the use of shared or acquired accounts outside official channels. For practical guides and security recommendations on key and access management, supplier support pages and industry good practice manuals are useful resources: Microsoft - MFA and authentication.

No less important is the human component: to inform and train employees about the dangers of using accounts of dubious origin, of sharing credentials or of using shortcuts that promise "premium access" through informal channels. The adoption of governance policies for the use of IA helps to clarify which tools are authorized, which data can be processed in public environments and which should remain in organization-controlled environments. Also, integrating underground market surveillance into corporate cyberintelligence programs allows you to detect leaks of accounts, keys or data before they are massively exploited; Flare offers monitoring services in that line, with examples of findings on your platform: example identified by Flare.

The evolution of the problem also poses regulatory and technical challenges for IA service providers. The developers of these platforms are expanding their efforts to detect and mitigate abuses, improve access controls and provide business capacities that better protect organizations. Still, effective security requires coordination between suppliers, security equipment and end-users and a mix of technical controls, clear policies and active monitoring of the digital ecosystem.

In substance, the emergence of an IA account market is a reminder of a classic security rule: any valuable service will feed incentives for its exploitation. The difference now is the magnitude and speed with which IA models multiply the operational capacity of malicious actors. If companies and institutions prioritize the protection of their access, educate their staff and adopt appropriate controls, they will be able to minimize much of that risk and take advantage of the benefits of the IA without turning their accounts into other goods on the illicit market.