In early 2023, ChatGPT shook the world, marking a major shift in artificial intelligence. Two years later, OpenAI continues its push toward artificial general intelligence (AGI), while tech giants like Meta have followed with powerful open-source models.
But even with the race to build more advanced AI, no single model can address every problem in every industry. Beyond the pursuit of bigger and better AI, models are being tailored to industries and integrated into specific applications and technologies.
Take the smart automotive industry, for example. Tesla’s Full Self-Driving (FSD) package, enhanced by vast datasets and AI models, now behaves more like a human driver than ever before.
But the synergy between AI models and the automotive industry goes beyond self-driving capabilities. AI is reshaping smart cockpits and vehicle safety too. In smart cockpits, AI has made voice commands feel more natural, intuitive—even lifelike. And that’s just scratching the surface of what AI could do for this space.
On the safety front, AI’s role is more subtle but equally crucial. As more cars become connected, security vulnerabilities and hacking threats have grown more sophisticated and widespread.
AI models can strengthen automotive safety by building smarter defense systems—something Callisto Technology, a Chinese automotive security company, is focused on. By using cloud-native, data-driven AI, Callisto aims to create a security system that continuously evolves to protect cars from modern threats.
In August, Callisto upgraded its automotive security AI model, adding a new set of intelligent agents with advanced capabilities. This AI now scans millions of logs in real time, identifying hacking attempts and anomalies, helping automakers stay ahead of security risks.
At the helm of Callisto is Yun Peng, a Baidu Apollo veteran and former chairman of Baidu’s security technology committee. When Baidu Apollo expertise intersects with advanced AI technology, what could it mean for the future of automotive security?
Baidu Apollo veterans delve into automotive security
Since Callisto’s founding in 2022, the company has taken a different approach to vehicle security. Instead of relying on in-car antivirus software, Callisto uses AI-powered cloud analysis and big data to monitor and prevent cyber threats and anomalies.
Internationally, Upstream Security has taken a similar approach, with backing from automakers like Volvo and Hyundai.
Callisto’s focus on AI stems from its founding team’s roots at Baidu. In 2017, Baidu’s then-president Lu Qi announced an “All in AI” strategy, drawing waves of AI talent to the company. But this surge also led to an exodus of top talent, with many launching their own startups. Baidu became known as an AI talent factory.
| Company | Founded | Founders (Former Baidu Staff) |
|---|---|---|
| 4Paradigm | 2014 | Dai Wenyuan, Chen Yuqiang |
| Nextdata | 2015 | Tang Huijun |
| NovuMind | 2015 | Wu Ren |
| Kuandeng | 2016 | Liu Ji |
| Moran | 2016 | Dai Shuaixiang |
| Pony.ai | 2016 | Peng Jun, Lou Tiancheng |
| WeRide | 2017 | Wang Jin, Han Shen |
| Mogo | 2017 | Zhu Lei |
| Trunk | 2017 | He Kan, Zhang Tianshi |
| DeepLearning.ai | 2017 | Andrew Ng |
| Allride | 2018 | Wang Jin |
| DeepRoute.ai | 2019 | Zhou Guang |
| Rino.ai | 2019 | Zhu Lei, Xia Tian |
| Haomo.ai | 2019 | Gu Wei |
| Senior | 2020 | He Kan |
| AutraTech | 2021 | Tao Heng |
| PhiGent Robotics | 2021 | Shan Jie, Pan Feng |
| Zelos | 2021 | Kong Hui |
| Callisto | 2022 | Yun Peng |
| Wuwen | 2022 | Liu Jiangang |
Much of this talent ended up in the automotive industry. Lou Tiancheng, once one of Baidu’s brightest programmers, co-founded Pony.ai, which is now valued at USD 8.5 billion. Other companies like WeRide, Haomo.AI, and Trunk have also thrived, thanks to Baidu alumni.
As AI models began to gain traction, these Baidu veterans saw new opportunities. Companies like 4Paradigm have already applied AI to various industry-specific challenges.
Callisto is part of this wave of Baidu talent venturing out.
Yun’s journey mirrors that of many Baidu veterans. He joined Baidu in 2014 and was soon put in charge of Apollo’s autonomous vehicle safety operations. By 2017, Yun’s team was responsible for safeguarding Baidu’s famous self-driving demo on Beijing’s 5th Ring Road. His work on Baidu’s autonomous vehicle security system eventually became part of the open-source Apollo platform.
After nearly two decades in the security industry, Yun realized that automotive safety was on the cusp of a new era.
“Just like internet security hit a turning point over a decade ago, the automotive industry is entering its golden age of cybersecurity. AI is going to be central to that transformation,” Yun said. Determined not to miss the opportunity, he convinced several Baidu colleagues to join him in founding Callisto.
The name “Callisto” comes from Jupiter’s second-largest moon, known for its stable surface despite being battered by meteor strikes. Scientists believe Callisto could serve as a base for deep space exploration, and Yun hopes his company will offer similar stability in automotive security.
After the AI model explosion in 2023, Yun became even more convinced that AI would be pivotal to automotive safety.
Just like Baidu founder Robin Li, Yun envisions AI-driven intelligent agents becoming the websites of the AI era, eventually creating a vast ecosystem of millions of potential agents.
“It’s like the internet boom in the early 2000s. Some websites faded away, but others became billion-dollar companies,” Yun explained.
Yun also believes AI models, especially in B2B industries like automotive safety, have the potential to create massive unicorns. Vertical AI models can solve problems that broader, generalized models cannot.
RELATED ARTICLE
Nipping automotive safety risks in the bud
The smart car industry has never been more dynamic. Vehicles are increasingly evolving into complex, robotic systems. But underneath the innovation lies a growing threat: cyberattacks.
In 2021, a hacker remotely accessed a Tesla Model S, exploiting vulnerabilities to take control of the vehicle, even shutting off its engine while it was in motion. Such incidents are becoming more common.
According to Callisto’s 2024 semi-annual automotive vulnerability and threat intelligence report, 70% of security incidents in the first half of the year were remote attacks, primarily targeting smart cockpits, charging stations, and autonomous driving systems.
The United Nations estimates that, by 2030, almost all vehicles will be internet-connected. Many of today’s connected cars already have up to 20 potential points of vulnerability, and as more features are added, the risks will multiply.
On social media, car owners frequently complain about false alarms from vehicle security systems, raising concerns about reliability.
Improving automotive safety requires a macro overhaul of security systems as well as micro-level optimization of user alerts. It’s a complex challenge for automakers.
China’s Ministry of Industry and Information Technology has introduced regulations to strengthen driver-assistance systems and offer guidelines for vehicle safety.
Meanwhile, Yun believes that automotive security needs to move toward proactive, self-improving systems. “Traditional defense methods are basic, but automakers need solutions that provide protection throughout a vehicle’s lifecycle, from R&D to daily operation.”
AI models can help automakers build these advanced safety systems. Callisto’s vehicle security operations center (VSOC) platform uses AI to analyze thousands of vehicle signals in real-time, distinguishing normal user behavior from potential threats.
Traditional security algorithms rely on preset rules, which can often trigger false alarms. For instance, if a car door is opened five times in a minute, the system might issue a warning. But what if it’s a family with kids who are frequently opening and closing the door?
AI models can learn from such behaviors and recognize when something is not a security risk, reducing unnecessary alerts.
Beyond technical integration, Callisto’s AI model can act as a conversational assistant, helping automakers manage compliance, safety, and data analysis throughout a vehicle’s lifecycle. This AI assistant can sift through hundreds of thousands of alerts and identify which ones are genuine threats.
During one client demonstration, Callisto’s AI model performed so well that the client asked to see the underlying code to verify that it was truly AI-driven. The team provided the proof on the spot.
Yun explained that AI models have eliminated the need for dedicated data analysts, boosting analysis efficiency by over 30%.
Callisto’s investment in AI has been extensive. The company trained its AI model on automotive regulations from over a dozen countries, using proprietary data from dealerships and product quality records. After a year of training, the model has made significant progress.
For automakers, building a secure system is a long-term effort. Callisto has reduced the cost of deploying private cloud security solutions, making it easier for automakers to adopt AI-powered safety systems.
KrASIA Connection features translated and adapted content that was originally published by 36Kr. This article was written by Xiao Xi for 36Kr.