An unknown individual recently posed as U.S. Secretary of State Marco Rubio and reached out to senior officials and dignitaries using AI-generated voice and text messages. The apparent goal, according to an internal State Department memo obtained by the Washington Post, was to get access to sensitive information or digital credentials.
The episode is a clear example of how AI impersonation is moving from theoretical concern into high-level communication channels. It also shows why messages that appear to come from trusted public figures now require careful verification before anyone responds, clicks, shares information, or accepts a new contact request.
What Happened
The attacker contacted at least five people while presenting as Rubio. The targets included three foreign ministers, a U.S. governor, and a member of Congress.
The impersonation used both voice and writing. According to the memo, the messages were generated with AI and were designed to closely resemble Rubio's speech and writing style. The attacker used Signal and text messages, and also adopted the identity "Marco.Rubio@state.gov," an address unrelated to Rubio.
The outreach did not follow one single format. In at least two cases, the attacker left voicemails. In another case, a Signal chat invitation was sent by text. The memo also says other State Department employees were approached by email.
Important details remain unknown. The State Department has not released the contents of the messages, the names of those targeted, or whether any attempt succeeded. The attacker's identity is also still unknown.
Why AI Voice Cloning Changes The Risk
Impersonation has long been part of social engineering. What AI changes is the level of realism that can be produced quickly, especially when the target believes a message is coming from a known authority figure.
Security researcher Hany Farid of the University of California, Berkeley, told the Washington Post that these attacks do not require advanced technology. Public figures often have many recordings available, and only short samples are needed to create a convincing voice clone.
According to the source article, audio samples lasting just 15 to 20 seconds can be enough to clone a public figure's voice through online services. An attacker can upload the sample, confirm they have permission to use the voice, and type the message they want delivered.
Voicemail is especially useful for this kind of fraud because it avoids live conversation. The attacker does not need to respond in real time, answer unexpected questions, or maintain the imitation under pressure. A short recorded message can still create urgency, familiarity, and trust.
The Communication Weak Point
The case also highlights a practical problem: senior officials and their teams depend on fast communication across many channels. That convenience can become a weakness when identity is assumed rather than verified.
The use of Signal was specifically noted in the source article as a concern raised in connection with government communications. The attacker used the encrypted messaging app and ordinary text messages to approach targets, which made the contact appear direct and personal.
For people receiving sensitive messages, the issue is not whether a platform is encrypted. The larger question is whether the person on the other end is truly who they claim to be. Encryption protects a channel, but it does not prove identity when a new invitation, voicemail, text, or email arrives unexpectedly.
The State Department has launched a full investigation and plans to introduce additional security measures. No further details about those measures were provided in the source article.
A Broader Pattern Of AI Impersonation
The Rubio impersonation attempt fits into a wider warning about AI-generated messages targeting public officials. The FBI previously warned in May about an ongoing campaign using AI-generated voice messages against senior U.S. officials and their contacts.
That campaign was described as an effort to steal information or money. The FBI advised people not to trust messages that seem to come from high-ranking officials unless they verify them first.
The source article also points to a similar incident last year, when AI-generated robocalls imitating President Biden were used in New Hampshire in an attempt to keep voters away from the primaries.
These examples show how the same underlying technique can be applied in different contexts:
- Impersonating a senior official to reach diplomats or elected officials.
- Using voice messages to target senior U.S. officials and their contacts.
- Deploying robocalls that imitate a political figure during an election process.
The common thread is trust. AI-generated voice and text can make a false message feel familiar enough that the recipient may act before verifying it.
What This Means For Verification
The most important lesson from the incident is not that every message is fake. It is that high-trust communication now needs a stronger habit of confirmation, especially when a request is unexpected or involves sensitive information, credentials, money, or access.
The FBI's advice, as described in the source, is straightforward: do not rely on appearances alone when a message seems to come from a high-ranking official. Verification should happen through a trusted route, not simply by replying to the same message or accepting the same invitation.
Open-source tools like Chatterbox are also lowering the barrier for voice cloning, according to the source article. That means the problem is not limited to well-funded attackers or highly specialized teams. If convincing impersonation becomes easier to produce, organizations will need to treat identity checks as part of normal communication rather than an emergency exception.
For government communication, the stakes are especially high because the targets may handle sensitive information and operate under time pressure. The Rubio case shows how an attacker can combine a familiar name, a realistic voice, a plausible writing style, and common messaging channels to create a serious security test.