Email and workplace cybersecurity startup Perception Point Inc. today announced the launch of a new artificial intelligence solution powered by OpenAI’s GPT-4 called GPThreat Hunter that can autonomously resolve complex cybersecurity issues when a case appears to be ambiguous.

Perception Point provides prevention-as-a-service for email, cloud collaboration apps and web browsers. Its service can be natively integrated for incident response for security teams to help reduce management overhead and improve user experience with insights into the threat landscape.

The company said that this new model builds on the accuracy of the company’s already existing successful capability of detecting and preventing 99.5% of threats against its customers, along with a dataset that includes years of human expertise in responding to incidents. By adding an automated response to its capabilities with an AI-driven solution, securities teams can stay on top of what they need instead of dealing with rote day-to-day tasks.

“By harnessing the power of the GPT-4 model, we are not only enhancing our detection capabilities but also elevating our stellar incident response services to unprecedented levels,” said Tal Zamir, chief technology officer at Perception Point. “Our customers will now benefit from an even more efficient incident response service in which our cyber experts can focus their resources on the most complex cases where human expertise is irreplaceable.”

GPThreat Hunter activates automatically when the company’s normal solution tripwires flag a situation as ambiguous. This makes it a perfect situation for AI to handle, which can then compile a detailed examination of the situation, complete with context, including evidence from the detection engine and all data points that led to the flag. The data is then interrogated by custom-trained large language models that have been trained by Perception Point and combined with GPT-4, which delivers an explanation of its determination.

The company said that this new multimodel approach combined with its years of knowledge and data taken from numerous threats is already better helping it protect customer data.

“By fusing the extensive knowledge, cyber research, and attack trend analysis of our cybersecurity experts with state-of-the-art gen AI models, we ensure that our customers are even more secure while receiving 24/7 service,” said Motti Elloul, vice president of customer success and incident response at Perception Point.

Perception Point has been hard at work elevating its capabilities with AI, including the release of GenAI Decoder, an LLM-based model for detecting email-based social engineering attacks crafted by generative AI. The company also released a solution that can detect a new form of email attacks that use Quick Response codes, called Quishing, designed to infect computers and devices with malware.

Although currently GPThreat Hunter only works on text, the company said that it intends to build multimodal analysis into the threat engine in the future that will allow it to handle images as well in the future. The company also plans to allow the agent to provide interactive question-and-answer reporting and real-world agent actions as well in a future update.

This article first appeared in Silicon Angle, written by Kyt Dotson on March 14, 2024.