Since the AI is helping hackers rapidly attack mass attacks, former Google security leaders have joined forces to build independent AI agents, which aims to prevent fashing, malware and business email agreements before reaching the user’s inboxes.
This is the mission AgencyA new email security startup, which has just emerged from stealth, has a $ 13 million seed financing in collaboration with Excel and Foundation Capital.
More than 90 % of successful cybrertax begin With a phishing email, CISA per US Federal CyberScurement Agency. A recently mobilized Straike Study (PDF) also found that fishing messages made by large language models (LLMS) have 54 % click throw rate in 2024, which is much higher than a 12 % rate for human written emails.
The purpose of the Education is to counter this growing threat with its independent AI agents.
Former Google Safe Browsing and Ricaptcha executives are founded by Sai Khoormaee and Ryan Lou, a start -up offers an orchestated network of real -time AI agents that inspects, analyzes and influences the risk of self -esteem, without relying on a particular set of specific rules. This approach challenges the general email security platform that relys on static rules and often requires extensive user training.
In an exclusive interview with the Tech Crunch, Khurmi said, “The collection of evil is associated with the PDF in an email. It is always where all the attacks started, and therefore I really wanted to solve the problem.”
Khurmi was the head of the product and director in Google until July 2023 in Google. At the same time, he met Lou for the first time, who spent almost a decade in Google and was part of the Safe browsing team.
He told Tech Crunch that Google experienced fashing technologies, a deep understanding of security from the company’s perspective, and how to develop and scale security business on the scale.
Prior to Google, Khurmi founded the Sales Intelligence Platform Cantastic, which was acquired by the Sugar CRM in 2016. Later, he served as a VP of product management for more than a year and a half before November 2024 before starting the Egisai.
Edisai has made reasoning agents, each of which is a custom built -in LLM according to a specific risk. Once the orchestating agent recognizes a danger or potential risk, he calls other agents in the network, which Khurmi has called “friends”. Then these agents analyze, argue with each other, and respond to the orchestrating agent with a decision.
Agents analyze the real -time analysis of each message component, including links, attachments, metadata, QR codes, and behaviors.
“From the construction of these tools in Google, what do we know about an email you need to analyze? What are all the sources of data? What are all techniques to attack, and what are the opponents of all the dirty things that we have been playing with these opponents for more than 10 years?” Khurmi said.
Although Egisai has currently created more than 10 agents for this task, Khurmi told Tech Crunch that there may be 50 to 100 agents over time as the opponents get better and try to fool the system.
He said, “I am fully convinced that in two years, the opponents will understand what we are doing. They will work again and attack what we are doing, and then we will need to build more agents to stay ahead of them.”
Unlike a common email security platform, which uses the rules -based approach, these AI agents have seen every possible form of these attacks a group of attacks and themselves automatically. Startups have developed numerous AI models that are created in accordance with various risks and specific industries, including venture capital and financial services.
According to Startup claims, compared to traditional solutions, agents of the Education agents, as well as detects threats, helps to reduce the incorrect positives by up to 90 %.
Google Workpace or Microsoft 365 Email account installing the Egisai system seems to be “no more than five minutes”, through an API, per scratch. Once setup, the startup will send a report in a couple of days, which is found in the environment with details, including false positive and false negatives. Then it will only go into the reading mode for a week and then activate the Singh.
Khurmi said, “It is so difficult to solve this very diverse problem in the email without this technology.
With Startups, San Francisco and New York offices, is currently running a pilot with users in the United States and Europe, and has already added three paying users, including data privacy compliance software loose and crypto payment platform mesh connect. Startup currently has a team of six members.
With the latest investment, Khurmi said that startup plans to enhance its technical skills and make a strong infrastructure to go to the market.