A Baltimore cybersecurity company is releasing a ChatGPT-style artificial intelligence to combat an expanding network of fake social media posts and threats from AI programs.
FoxGPT is a language model similar to the famous OpenAI program ChatGPT. Language models are “trained” with a certain dataset, like news articles or public documents, and then can answer questions and prompts based on that data. The Baltimore public company (Nasdaq: ZFOX) announced the new product on May 10 and is going to slowly release the AI tool to customers throughout the year as part of the company's cybersecurity platform.
While more mainstream language models like ChatGPT are trained using internet articles or artworks to replicate visuals, FoxGPT is trained on cybersecurity information to determine if an email or other piece of information is malicious, ZeroFox Chief Technology Officer Mike Price said. For example, if ZeroFox inputs the content of a tweet into FoxGPT the language model would determine whether or not the tweet is likely to be malicious or a bot by seeing if the content of the prompt is similar to other examples of cyber threats. The chatbot is supposed to supplement ZeroFox's workflow, speeding up the processing of the data ZeroFox receives from clients to help identify problems quickly.
The debut of FoxGPT comes amid a recent explosion in AI that has sent shockwaves through a variety of different industries. The use of AI products is one of the linchpins for the current Hollywood writer’s strike since many people in creative professions worry that ChatGPT could be used to write movie scripts or provide commercial art in a way that could put real workers out of a job. Morgan State University is conducting research into the ethical use of AI through a federally funded research center.
Generative AI is also creating problems in the cybersecurity world. AI makes common cyber attacks like phishing emails look more believable or creates deep fakes by creating videos, pictures, or writings in the style of a brand or famous person. A picture of Pope Francis in a puffy jacket is one of several deepfakes, or images edited to resemble fake events, that have gone viral in the past several months. The "Balenciaga Pope" fake image is a more harmless example of how AI tools can fool many people and spread rapidly on social media as people share strange pictures of celebrities or false information. Price believes that the upcoming 2024 election season could be rife with fake photos on video spreading political misinformation on social media.
“Just as companies have leaped to embrace advances in generative AI in the past few months, threat actors are doing the same thing – it has to be taken seriously and cybersecurity companies have to adapt to the changing threat landscape, which is exactly what we’re doing with the rollout of FoxGPT,” Price said.
The AI program is not the first time ZeroFox has worked to keep up with the latest technological and economic trends. CEO James Foster told the Baltimore Business Journal in March that he viewed the collapse of Silicon Valley Bank as an opportunity to grow the business because the volume of startups changing banking information created an environment ripe for fraud.
ZeroFox's stock, like that of many companies that have gone public through a Special Purpose Acquisition Company (SPAC) merger, has struggled, declining by over 90% to 83 cents per share since the company went public last August. The company lost $16.4 million in its fiscal fourth quarter ending Jan. 31 and saw revenue of $45.3 million. Foster previously told the BBJ he expects the company to reach a break-even cash flow late next year. ZeroFox recently acquired LookingGlass Cyber Solutions, a cybersecurity company, for $26 million.
