Ahead of the US midterm elections, federal agencies have put election offices and political groups on notice about cybersecurity risks, including AI-generated phishing attempts designed to deceive recipients.

On 16 October 2026, the FBI and the Cybersecurity and Infrastructure Security Agency (CISA) jointly published a notice stating that artificial intelligence tools have the potential to be exploited by threat actors to craft deceptive emails and text messages that are harder for recipients to detect as fraudulent. Such communications are capable of harvesting login credentials, extracting private data, or opening a path into accounts held by those running elections and political campaigns.

Timed to land before November's midterm elections, the alert targets political organisations and election offices, both of which are expected to draw heightened attention from cyber actors during that period.

Automated writing tools have lowered the barrier for those with fraudulent intent, enabling the production of text that reads as polished and carries an air of credibility. Deceptive messages — whether arriving as emails dressed in the formal tone of official notices or as texts that appear to originate from a known contact or service — are designed to pass as legitimate. The federal alert noted that these tools could be leveraged to make phishing and deception attempts against election officials, candidates, and political organisations more sophisticated.

Recipients of a convincingly crafted fraudulent message may act on embedded links, transfer login credentials, or disclose personal information that was intended to stay private. One point of focus in the alert is the pace at which those with malicious intent can generate personalised messages aimed at distinct individuals.

Coordinating electoral processes draws on both protected data streams and interconnected digital infrastructure. Campaign operations routinely store voter records, staff correspondence, and login credentials — a combination that draws the attention of those looking to exploit digital vulnerabilities. When an email account is breached, private messages may become accessible to an unauthorised party, who can also use the trusted address to dispatch deceptive communications to others.

Disruptions to day-to-day operations, along with the potential for widespread confusion, remain a real possibility at a time when stability matters most. Voting offices and campaign staff are advised by the alert to treat any out-of-the-ordinary request for login credentials or protected data with scepticism, even when the communication carrying that request appears entirely routine.

No compromise to the election process itself has been indicated by the alert. The underlying concern centres on AI's capacity to sharpen the potency of well-established cyberattack methods at a point when political organisations face heightened exposure to those threats.