With the rapid development of artificial intelligence (AI) and large language model technologies, they have brought unprecedented convenience and transformation to people’s lives. From the explosive popularity of applications like DeepSeek to the accelerated adoption of edge AI devices such as AI PCs, AI is everywhere. However, the misuse of AI technologies has also created new security threats. Among them, AI-powered deepfake scams—highly covert and harmful—are increasingly becoming a major challenge in the field of cybersecurity.
During the Two Sessions, representatives of the National People’s Congress and members of the Chinese People’s Political Consultative Conference called for legislation to regulate AI-based “face-swapping and voice-mimicking” technologies, sparking widespread attention. At its core, AI face-swapping is an innovative application of generative AI that uses deep learning to produce highly realistic fake videos or images. As the technology matures, forged media has become indistinguishable from reality, giving criminals opportunities to commit fraud, identity theft, and the dissemination of disinformation, causing serious damage to individuals, businesses, and society as a whole.
To address the growing threats posed by malicious AI applications such as deepfakes, Lenovo Group, leveraging its deep expertise and keen insights in AI, is set to launch a groundbreaking anti-fraud technology—the “Deepfake Detection Engine.” In Lenovo’s newly released promotional video, “Using AI to Protect AI,” more details of the technology are showcased. It can accurately identify various forms of deepfake forgeries and effectively curb the spread of AI-driven fraud. The detection engine reportedly achieves an accuracy rate of up to 96%.
Among many AI face-swapping technologies, Deepfake is undoubtedly one of the most mature and representative. The term “Deepfake,” derived from “deep learning” and “fake,” concisely captures its essence. As an important subfield of AI deep learning, deepfake technology has attracted great attention since its inception. However, it has quickly moved from laboratories to mainstream society, appearing frequently in applications and triggering widespread public concern.
Built on the open-source DeepSeek large model, Lenovo’s Deepfake Detection Engine reflects the company’s precise grasp of cutting-edge technologies and effective utilization of them. It can also be deployed locally on AI PCs, further strengthening Lenovo AI PCs’ protection of personal data and privacy. With the powerful performance and broad applicability of DeepSeek, the detection engine has a solid technical foundation. Lenovo’s R&D team continues to explore content recognition and anti-fraud technologies, optimizing algorithms and models to build an efficient, reliable AI security tool. Meanwhile, Lenovo’s anti-fraud initiative has already attracted support from Sichuan Anti-Fraud Center, Tencent Cloud, and Qi An Xin, among others.
The deepfake detection method proposed by Lenovo has several technological advantages. From a technical perspective, it leverages advanced deep learning algorithms to analyze subtle details in images and videos. By constructing complex neural network models, the detector learns the fine differences between authentic human faces and deepfakes, enabling accurate detection. Compared with traditional rule-based methods, this data-driven approach is more precise and adaptive, better handling the evolving nature of deepfake technologies.
In terms of functionality, Lenovo’s technology has achieved significant results. For example, in online video conferencing, the system can detect participants’ video feeds in real-time. If a potential deepfake is detected, it immediately alerts the user, enhancing meeting security. This function is especially critical for businesses during negotiations or confidential exchanges, where verifying identity is paramount. On social media, the technology can automatically scan videos and flag suspected deepfakes, helping ordinary users identify misleading content and maintaining healthier online ecosystems. When using AI PCs, the system provides real-time monitoring of deepfake threats and issues timely warnings, offering users comprehensive protection across work, entertainment, and daily activities.
The technology is highly adaptable, supporting a wide range of devices including laptops, desktops, tablets, and smartphones. This cross-platform compatibility demonstrates Lenovo’s user-centric design philosophy, ensuring that wherever users are and whatever smart device they use, they benefit from Lenovo’s AI-powered security protection.
In addition to its accuracy and wide applicability, Lenovo’s detection algorithm is also highly efficient. It can analyze and return results for images or videos within just five seconds, offering automated analysis and real-time feedback. In today’s fast-paced information age, speed is critical. Quick detection results provide users with valuable response time to mitigate potential fraud and losses.
Furthermore, the technology supports iterative upgrades with strong defensive capabilities. As deepfake attacks evolve, with new technologies like Faceswap, Face2Face, and NeuralTextures emerging, Lenovo’s detection engine continuously collects new samples to retrain and optimize its models, maintaining ongoing system updates. This dynamic, adaptive defense mechanism ensures the detector remains at the forefront of the industry, providing long-term effective protection.
The launch of Lenovo’s Deepfake Detection Engine is not only a technological innovation but also an embodiment of the company’s corporate social responsibility. Lenovo adheres to its philosophy of “Human-Centered Intelligence,” which calls for re-examining AI technologies and their impacts from a people-first perspective. This means prioritizing human needs and well-being in AI R&D, design, application, and interactions. While AI development introduces new risks, proactive innovation can also provide powerful safeguards against them.
From the perspective of user protection, Lenovo’s technology provides a solid security barrier. In daily life, users may suffer financial losses or personal data breaches due to AI fraud. With the Deepfake Detection Engine, users can use smart devices and online services with greater peace of mind, free from the threat of AI-driven scams. This enhances both their quality of life and their trust in technology.
For businesses, the technology helps foster a fairer, more secure commercial environment. Verifying identities and ensuring authenticity are vital in commercial activities. By accurately detecting deepfakes, companies can effectively prevent fraud, protect trade secrets, and reduce risks, supporting healthier business development and market order.
At the societal level, Lenovo’s anti-fraud technology contributes to social stability and harmony. The prevalence of AI fraud not only causes losses for individuals and enterprises but also risks triggering a broader crisis of trust. By promoting and applying advanced anti-fraud technologies, society can curb fraudulent activities, boost public confidence in technology, and promote sustainable social development.
Jia Chaohui, Senior Vice President of Lenovo Group and General Manager of the Global Innovation Center at the Intelligent Devices Business Group, stated: “In an era when AI face-swapping scams are rampant, the Lenovo Global Security Lab team is building the ‘AI Demon Mirror’ based on the DeepSeek open-source model, dedicated to exposing deceptive AI-generated content. Anti-fraud is essentially a tug-of-war between offense and defense—while forgery technologies evolve daily, engineers work tirelessly around the clock. Victory cannot be achieved alone. Lenovo looks forward to joining forces with more technology partners to build a digital security defense line. We firmly believe that AI is not a hidden pair of eyes lurking in the shadows, but a light that should truly illuminate life.”
Video：https://www.youtube.com/watch?v=NeupJrbgmgM