Dream Raises $100M to Strengthen AI-Driven National Security

A startup led by the ex-CEO of Israeli spyware maker NSO Group raised $100 million on a $1.1 billion valuation to strengthen national cybersecurity through predictive intelligence and real-time protection.

See Also: New Attacks. Skyrocketing Costs. The True Cost of a Security Breach.

Tel Aviv, Israel-based Dream will use the Series B funding to more effectively secure governments and nations through AI-driven, real-time detection and holistic protection across IT, OT, legacy systems and cloud environments, according to co-founder and CTO Gil Dolev. The proceeds will go directly toward expanding Dream’s proprietary cyber language model to enhance predictive threat detection, he said.

“The idea of raising money now is because of our growth,” Dolev told Information Security Media Group. “We actually added $30 million in sales to governments and nations and organizations. There is a huge demand for that. The increasing sophistication of AI-driven cyberthreats makes it essential for us to keep scaling and to create a global sales market right now.”

Dream, founded in 2022, employs 159 people and previously closed a $33.6 million Series A round in November 2023 led by Aleph and Group 11. Dream has been led since its inception by Shalev Hulio, who co-founded spyware firm NSO Group in 2009 and served as its CEO from 2019 to 2022, leaving amid allegations that NSO’s tools were misused by governments and other agencies to hack mobile phones (see: Seeking a Buyer, NSO Group Announces Fresh CEO Plus Layoffs).

Dream’s decision to raise Series B was driven by the company’s rapid growth and increasing demand for AI-driven cybersecurity solutions, particularly from governments and nations, Dolev said the Bain-led round is crucial for accelerating Dream’s R&D efforts, particularly an expansion of the company’s cyber language model capabilities with a focus on hardware investment for larger-scale model training.

Dolev said Dream wants to expand the AI model “and make it be able to answer much more sophisticated prompts,” Dolev said. “We want to increase it and to make it much bigger. We want to invest the money in getting better hardware and better capability to train it on bigger parameters.”

What Sets Dream’s Cyber Language Model Apart from Other LLMs

Dolev said Bain’s expertise in scaling technology companies and expertise with cybersecurity investment were key factors in selecting the firm as lead investors. He is excited to have Bain Partner and former Symantec CEO Enrique Salem join Dream’s board. Dolev views Salem as a strategic advisor and mentor who will bring critical industry insights on how to scale a cybersecurity company globally.

“Bain is one of the most appreciated and the biggest investors in the world and in the United States,” Dolev said. “Their deep expertise in taking and scaling companies was a big factor. The most important for me is the relationship with Enrique Salem. To partner with people like Enrique is like having a global security strategist that can help us as a team take this company forward and scale up.”

Unlike enterprises, which typically deal with targeted attacks from cybercriminals, nations must defend against millions of attackers, including state-sponsored adversaries using zero-day exploits and AI-powered attacks, Dolev said. He sees national cybersecurity as primarily a data challenge, where success depends on processing and analyzing massive datasets to detect and mitigate threats effectively.

“The threats against nations are huge,” Dolev said. “To increase the resilience of the nation, you must deal with a lot of data.”

Unlike general purpose large language models, Dream’s cyber language model is trained on datasets of cyberthreat intelligence and is capable of automating complex cybersecurity tasks such as threat detection, prediction and response. Dream plans to scale the model to 10 times its current size, which will increase its capability to process more extensive datasets and deliver real-time insights, Dolev said.

To scale the model, Dolev said Dream will invest heavily in data collection and model training, expanding the proprietary database with contributions from security researchers. Dream also plans to acquire high-performance computing infrastructure to train its large-scale models efficiently. Dolev said Dream also will create multiple foundation models within the architecture, each tailored for different cyber issues.

“This data is not just based on any open-source data,” Dolev said. “It’s based on our own knowledge, our own experience with this type of thing. The idea of a CLM is to combine between the attacker and defender. So our CLM doesn’t think like a defender only, but as an attacker as well, to be able to extend and predict what is going on.”

Geographic, Partnership Expansion on Dream’s To-Do List

Forming partnerships with cloud service providers and AI hardware leaders will help Dream integrate its technology into critical digital infrastructure, enhancing market reach and providing value to mutual customers, Dolev said. These partnerships are crucial for protecting national-scale digital environments and expanding Dream’s presence in key industries, according to Dolev.

Dream has achieved strong traction in Europe, the Middle East and Southeast Asia, which Dolev said are their primary revenue-generated regions. Dolev said the company waited to approach the United States market until they were confident in the product’s maturity and Dream’s ability to compete effectively.

“The reason we want now to go to the United States, and not before, is because we wanted to come as strongly as we can, as big as we can, to this market,” Dolev said. “This market is a great opportunity and a great option for Dream to scale itself and scale its impact on nations.”

Dream wants to increase its footprint in new geographic regions, secure additional contracts with national governments, and scale its cyber language model to handle real-time threat intelligence at larger scale. Dolev said a key component of Dream’s success will be achieving more autonomous, AI-driven security actions that reduce human dependency and increase protection speed and efficiency.

“If something I really believe is the ability that we can redefine national cybersecurity, which means expanding into new government markets, expanding into nations, and new geographic regions,” Dolev said. “And on the other side, it’s to increase the CLM capacity to process and correlate in real time and spread intelligence on a bigger and bigger scale.”