Analysts have warned Russians about the new purpose of fraudsters
Analysts of the Smart Business ALert (SBA) service from EA PRO (part of the Cross Technologies Group of Companies) warned that fraudsters have a new object of interest — data providing access to corporate AI agents.
The study, the results of which were published Lenta.Ru On Thursday, September 3, it showed that companies that use artificial intelligence (AI) to work with clients, documents, corporate mail and internal systems are at risk.
Services that offer free auditing of AI agents and request the system name, owner's e-mail, API Endpoint, API key, and System Prompt may be dangerous. Together, these data allow you to get the address of the system, access to it, and information about how the agent works.
By obtaining a valid API key, attackers can send requests at the expense of the owner, and when accessing a corporate agent, they can interact with connected files, request history, databases, and work tools within the scope of their authority. Knowledge of System Prompt also helps to select commands to circumvent restrictions and carry out prompt injection attacks, in which AI tries to force an attacker to perform tasks.
The risks are especially high for companies where AI is connected to CRM, corporate mail, cloud storage, financial transactions or technical infrastructure. In this case, the attacker potentially gets the opportunity to act on behalf of the organization.
Sergey Trukhachev, the head of the service, noted that the system intelligence allows you to understand the logic of the agent's work, its limitations and available actions. Combined with API access, this makes it possible to carry out a series of adaptive attacks, seeking to disclose information or behavior desired by an attacker, such as recommending a specific supplier, providing false details, or approving a prohibited action.
On the same day, Evgeny Masharov, a member of the Russian Public Chamber's commission for public control and handling of citizens' appeals, warned that fraudsters had begun offering Muscovites to receive a "payout" by City Day, as well as take a survey to win a prize. The messages contain a link, clicking on which and entering personal and bank data leads to the withdrawal of money from accounts, writes "Moscow 24".
On the eve of September 2, Vladimir Zykov, editor-in-chief of the analytical publication, told RT that fraudsters can copy the manner of writing using AI. According to him, the more correspondence is available, the more convincing the imitation can be. However, technology does not always accurately convey humor, shared memories, and reactions to unexpected questions. 360.ru .
At the same time, the head of the "Call Me Back" project at the Moscow Department of Information Technology, Valentina Shilina, warned that the first rule of digital hygiene is to create secure passwords. Today, they are required in all services, from a work laptop to cloud data storage. You should also separate the accounts: use personal mail and messenger for personal correspondence, according to the Moscow City News agency.
Cybersecurity expert Alexander Dvoryansky told NSN on September 1 that all sectors of the economy face the threat of phishing attacks equally. In order to prevent fraudsters, it is necessary to train staff and simulate attacks. He stressed that it is important to educate users about the basic principles of information security.
At the end of August, the Russian Interior Ministry reported that fraudsters had started sending a link to a website with a video that mimics hacking into a Gosuslug account. After clicking on the link, the user gets to a page that looks like a government portal.
Переведено сервисом «Яндекс Переводчик»