Europol warns that a criminal breach of ChatGPT is coming

From phishing to disinformation and malwareIn a new report, Europol says the rapidly evolving capabilities of chatbots will not only be used to improve humanity, but also to deceive it.

the chatcreated by US startup OpenAI, appeared in November and was quickly won over by users who were amazed at its ability to clearly answer difficult questions, write sonnets or ciphers, and even pass exams.

“The potential exploitation of AI systems by criminals presents a grim prospect,” said The Hague-based Europol.

She said Europol’s new “innovation lab” looked at the use of chatbots as a whole, but focused on ChatGPT during a series of workshops because it is the most prominent and widely used.

The agency found that criminals can use ChatGPT to “significantly speed up the investigation process” in areas they know nothing about.

It added that this could include writing to commit fraud or providing information on “how to break into a home through terrorism, cybercrime and child sexual abuse”.

It said the chatbot’s ability to mimic speech style made it particularly effective at phishing, where it lures users into clicking on fake email links that then attempt to steal their data.

ChatGPT’s ability to quickly produce original audio texts makes it “ideal for advertising and disinformation purposes, allowing users to produce and disseminate messages that reflect a given narrative with relatively little effort.”

ChatGPT can also be used to write computer code, especially for criminals without technical knowledge. Europol said: “Europol has nothing to do with this technology.”

“This type of automated code generation is particularly useful for those criminals who have little or no knowledge of programming and development,” he said.

Europol said an early study by US-Israeli cyber threat intelligence firm Check Point Research (CPR) showed how a chatbot could be used to infiltrate online systems by creating phishing emails.

Europol said that while ChatGPT has safeguards, including content moderation, that it won’t answer questions deemed harmful or biased, it can be bypassed with clever prompts.

He added that artificial intelligence is still in its infancy and its capabilities are expected to improve over time.

“It is extremely important to raise awareness of this issue to ensure that any potential vulnerabilities are discovered and closed as quickly as possible,” Europol said.

Source: Agence France-Presse