Artificial intelligence chatbots and large language models (LLMs) present a rising security threat, the UK’s intelligence agency has warned.
The National Cyber Security Centre (NCSC) issued the warning on Tuesday, advising people to “take great care” with data they choose to submit to chatbots, given companies will “almost certainly” access it.
It added sensitive information should not be put into AI products and that people were best to avoid telling chatbot’s information they would not want in the hands of the public.
NCSC’s concerns were prompted by the record-breaking rise of OpenAI’s ChatGPT, which gained 100mln users in just two months after its November launch, leading a host of other firms to invest in similar programmes.
Microsoft, having invested US$10bn into OpenAI, was joined by Google owner Alphabet Inc (NASDAQ:GOOG) in announcing new chatbots last month, with the craze spreading to Salesforce’s Slack messaging service last week.
Microsoft subsequently admitted staff were reading users’ conversations with its AI bot, which called itself Sydney, attempting to seek out “inappropriate behaviour”.
“There are undoubtedly risks involved in the unfettered use of public LLMs,” NCSC added, “individuals and organisations should take great care with the data they choose to submit in prompts”.
Discussing other risks presented by the emerging tech, NCSC claimed criminals could use AI to write phishing emails and to come up with new scamming techniques, as people were now able to easily generate messages in other languages.
"Loose lips sink ships," GCSC said, referencing a Second World War US saying, implying "loose prompts" should be thought of in the same way.