Attackers abuse shared ChatGPT and Claude chats to spread malware

What happened

Attackers have started exploiting the chat-sharing features in ChatGPT and Claude to distribute malware. By creating chats that resemble error messages or installation guides, these malicious actors share conversation URLs that lead users to download malware. Because these chats are hosted on the official, trusted domains of the AI platforms, traditional security tools often miss them, allowing the malware to spread under the radar.

The risk

This tactic takes advantage of the trust users place in ChatGPT and Claude’s domains, numbers trusted by most corporate firewalls and endpoint protections. Since the malicious content is embedded in a shared chat rather than a separate website or attachment, it bypasses many standard defenses. This raises the risk for businesses, IT teams, and individual users who rely on these platforms for safe collaboration and information exchange. The attack also complicates detection because the initial vector does not resemble classic phishing or malware delivery methods.

Why it matters

Attackers using legitimate, trusted domains weaken a key assumption of many security models—that trusted domains are safe. This forces security teams to reconsider how they handle URLs and files originating from well-known AI service providers. For builders and product operators, it pressures a review of shared chat controls and content validation. For enterprises and end users, the need grows for enhanced vigilance and possibly new security policies to block or test links from shared AI chats. The move also shows that criminals are adapting quickly to AI platforms’ features, turning collaboration tools into malware channels.

Who should pay attention

IT and security teams must pay close attention, as existing perimeter defenses are unlikely to catch this new attack vector. Developers working on AI chat and collaboration platforms should consider adding safeguards against embedding dangerous content in shared conversations. Businesses and users must exercise extra caution before clicking links in shared AI chats, even if they appear on official domains. Investors and regulators tracking AI platform risks should note this shift as a signal that new controls on sharing and content moderation may become necessary.

What to watch next

Expect AI chat platforms to strengthen protections on shared chats, possibly limiting types of content or introducing automatic scanning to block malicious payloads. Security vendors may develop specialized tools that monitor URL reputation and content within trusted domain chats. Users and enterprises should track how policy and training evolve to include this new threat vector. The broader battle over AI platform security will intensify as attackers continually look for novel ways to exploit collaboration features.

AI Quick Briefs Editorial Desk