Security News

Cybercriminals' Overreliance on AI Is Producing Sloppy Attacks, Proofpoint Researcher Says

Infosecurity Magazine · 13 Nov 2026
Key Takeaway Keep training staff to question unexpected emails and login pages, because mismatched subjects, blank messages and generic-looking websites are useful warning signs, though attackers may improve over time.

Since 2023, businesses have been warned that AI would make attackers better, faster and stronger. Speaking at the IRISSCON 2026 conference in Dublin on November 11, Proofpoint principal threat researcher Selena Larson challenged that view. She called it "marketing nonsense" and said the data does not back it up. AI has helped threat actors increase the speed and scale of their campaigns, she said, but some groups have become so reliant on it that they make frequent errors, which she described as "AI slop".

Proofpoint has found AI-generated malware that simply does not work. Larson said the code often includes comments describing one function while the malware does something completely different. Phishing campaigns show similar problems. She highlighted TA4903, a group focused on business email compromise and phishing, which has historically impersonated US government entities with sophisticated lures. Since adopting AI tools, its lures have declined in quality. Emails often do not match their subjects or attachments, and some arrive with a blank message body.

The group also now mainly builds phishing websites with AI coding tools. These sites tend to be simple and share similar colours and formatting, which makes them easier for targets to spot. Larson added that using AI to get region-specific grammar and spelling right can even make lures less believable.

Summarised by CISO AI from Infosecurity Magazine, written with Claude Sonnet 5.5. We link back to every original so you can read it yourself.