HydroX AI, a large model security and safe AI technology company, has raised $4 million in angel funding — led by Oasis Capital

HydroX AI recently closed a $4 million angel round led by Oasis Capital, with MiraclePlus and Atom Capital participating. The funding will go toward product R&D, market expansion, and team growth. An investment lead at Oasis Capital said: "The broad integration of AI and society is just beginning. The safety issues here go far beyond privacy and infiltration — they touch on social ethics, knowledge structures, and cognitive influence. In the future, safety won't just be a protective measure; it will be essential infrastructure that evolves alongside AI itself. We're thrilled to explore social safety and ethical responsibility in the age of AI alongside a world-class team with deep expertise in security and privacy." ZL, founder and CEO of HydroX AI, said: "Over the past year, Oasis has been one of the most professional and forward-thinking firms we've engaged with. Riding this massive wave of large language models, we're excited to work hand in hand with Oasis to meet an era full of both opportunity and challenge."

Founded in July 2023 in California, HydroX AI is a tech company focused on AI Safety and Safe AI, dedicated to providing comprehensive, multi-layered AI security solutions for LLM companies, enterprises deploying LLMs, and other B2B sectors. In today's rapidly evolving LLM landscape, model safety is an unavoidable priority. AI Safety focuses on the models themselves, breaking down into alignment, interpretability, robustness, and other issues. Put simply, it means putting LLMs through rigorous "quality control" to ensure they are safe and won't produce severe consequences or harm society. HydroX AI founder and CEO ZL said that compared to traditional cybersecurity, AI Safety is a relatively new field with more possibilities and commercial opportunities. ZL believes that LLMs are essentially black boxes, and such black boxes invite novel attack methods — prompt injection, for example. Many open-source AI projects have already been found to contain compromised code, and publicly available training datasets can also create security vulnerabilities. These challenges demand new, targeted solutions. AI Safety primarily depends on the LLM training process, and some related technologies have already emerged. Major LLM companies like OpenAI release corresponding safety techniques when launching models; NVIDIA has developed specialized guardrails to vet LLM outputs; and some universities have designed safety evaluation frameworks to assess model security. HydroX AI's focus is on risk control and data protection for LLMs. ZL said these are currently the most urgent and widely needed areas within AI Safety. If a model is injected with malicious code or scripts, users could end up having their account passwords, bank card numbers, and other personal information stolen after interacting with it. HydroX AI will offer customers a one-stop solution. Specifically, the company is building a standalone platform that creates an end-to-end closed loop from problem detection to resolution to continuous monitoring. Security is a long-term dynamic battle — no matter how strong the defense, stronger attacks will follow. HydroX AI is continuously expanding its detection library to make problem identification more comprehensive and accurate, improving the platform's ability to spot issues. Currently, HydroX AI has launched a free product on its website, mainly for new users to experience and raise awareness of this emerging field. ZL revealed that the company will roll out its first-generation one-stop platform by the end of this year. On the business model front, HydroX AI will start with customized development for clients to build benchmark cases, then gradually transition to a SaaS model with subscription fees. Given current staffing and resources, the company is initially targeting AI companies building foundation models at home and abroad, as well as model companies in healthcare and social media, to deliver one-stop solutions for these customers. Founder ZL previously worked at ByteDance, Meta, and LinkedIn, bringing over a decade of experience in cybersecurity. The core team has six members, all from Google, Meta, LinkedIn, and other major tech companies. Source: 36Kr Intelligent Emergence, author Lin Weixin Sponsor: Vitality

What do you think vitality is?

Vitality is not just drive — it's the resolve to safeguard the digital world.

—— ZL, Founder & CEO of HydroX AI*