AI, ML and automation have grown rapidly in recent years. Increasing use of smartphones, laptops, IoTs, and other devices has made lives easers and made business more efficient and agile. This has led to an increase in productivity and efficiency, but it also comes with its challenges. Many of us generate texts from AI-enabled sources, research report shows that AI-generated texts could increase people’s exposure to threats.
According to WithSecure research, large language models are susceptible to abuse through creative prompt engineering, pushing humans to become even more skeptical about what they read. On the other hand, nearly universal access to models that deliver human-sounding text in seconds presents a turning point in human history.
“The fact that anyone with an internet connection can now access powerful large language models has one very practical consequence: it’s now reasonable to assume any new communication you receive may have been written with the help of a robot,” said WithSecure Intelligence Researcher Andy Patel.
“Going forward, AI’s use to generate both harmful and useful content will require detection strategies capable of understanding the meaning and purpose of written content,” said Patel
The report shows a series of experiments performed using GPT-3 (Generative Pre-trained Transformer 3)–language models which use ML to generate text.
Multiple experiments assessed how input converted to the currently available models affected the synthetic text output. The goal was to determine how AI language generation can be misused through malicious and creative prompt engineering in hopes that the research could be used to direct the creation of safer large language models in the future.
The experiments covered phishing and spear-phishing, harassment, social validation for scams, the appropriation of a written style, deliberately divisive opinions, using the models to create prompts for malicious text, and fake news.
The model’s responses in these use cases, along with the general development of GPT-3 models, led the researchers to several conclusions, including the following four (but not limited to):
- Prompt engineering will develop as a discipline, as well malicious prompt creation.
- Adversaries will develop abilities enabled by large language models in unanticipated ways.
- Identifying malicious or abusive content will become more difficult for platform providers.
- Large language models already allow criminals to make any targeted communication as part of an attack more effective.
“We began this research before ChatGPT made GPT-3 technology available to everyone,” Patel said. “This development increased our urgency and efforts. Because, to some degree, we are all Blade Runners now, trying to figure out if the intelligence we’re dealing with is ‘real,’ or artificial.”
Also Read: Slack Security Concerns and DLP solutions
