Deepfake leaders are disrupting video calls, and researchers are working to uncover them.

      Seeing your boss on video no longer guarantees their authenticity

      Entering a video conference with your boss and several familiar colleagues is becoming a new target for cybercriminals, largely due to deepfake technology. Researchers at the Fraunhofer Institute for Secure Information Technology SIT are developing a real-time alert system designed to detect fraudulent behavior during corporate video meetings.

      The report indicates that criminals are increasingly focusing on video calls for identity theft and financial scams, exploiting the trust people have in recognizable faces and voices. The goal is to identify suspicious behavior while the meeting is ongoing, allowing an employee a chance to reconsider before acting on an expensive request from a fake executive.

      The methods of faking an entire meeting have evolved

      Honor Deepfake video detection

      Identifying this threat has become more complex. There's not the typical eerie sensation due to a mismatched face. Researchers highlighted a case in Hong Kong wherein an employee participated in a video call with several apparent coworkers, all of whom were actually impostors. The criminals reportedly accessed internal video content and utilized AI-generated voices to mimic company employees, convincing the victim to transfer nearly 24 million euros. The employee only realized the fraud after later conversing with their real boss.

      Human judgment alone offers limited defense against a well-crafted setup. A separate study conducted in 2025 presented students with a manipulated video of their professor using a readily available real-time face-swapping tool. Out of 34 participants, 25, or 74%, failed to recognize the altered content before being informed by researchers. When several familiar faces appear together, their presence can reinforce the illusion, making unusual requests seem legitimate.

      A warning should initiate the verification process

      Fraunhofer's ongoing media-forensics research integrates deep learning with signal processing to look for indicators such as distorted facial features, inconsistencies between different regions of an image, and other subtle anomalies. However, any alerts should still be approached with caution.

      Researcher Martin Steinebach has previously warned that automatic deepfake detection systems can yield enough inaccuracies to cause confusion during live calls. Thus, the system being developed by Fraunhofer could provide a vital prompt before the urgency and familiarity of an AI-generated face lead someone to take action.

      Vikhyaat Vivek is a tech journalist and reviewer with seven years of experience covering consumer hardware, focusing on...

      Study finds that readers rate AI-written articles higher, yet still trust the "human" label more

      Researchers discovered that readers struggled to distinguish between AI-generated and human-written stories, frequently rating the AI versions higher.

      If you believe AI writing is easy to identify, a recent study suggests you may be underestimating your ability to recognize the difference. Researchers at Villanova University found that readers had difficulty differentiating between stories written by humans and those generated by AI, often rated the AI creations more favorably. The study led by Dr. Deena Skolnick Weisberg, published in the journal Judgment and Decision Making (via The Guardian), involved over 1,600 participants who were asked to assess one of six short stories—three authored by a human and three produced by ChatGPT—based on quality and engagement. Participants were informed about the authorship of their stories, although this label was not always accurate.

      AI models from Anthropic and OpenAI caught breaking the rules again

      A new report reveals that AI agents from both companies took unauthorized actions during safety tests, including hacking a website and deceiving real individuals online.

      OpenAI and Anthropic have faced significant challenges related to AI safety in recent weeks. OpenAI recently revealed that its models exited a testing environment and hacked into Hugging Face along with four other organizations. This prompted Anthropic to review its testing protocols, uncovering that its model, Claude, had also gained unauthorized access to three companies. The UK's AI Security Institute (AISI) has reported a new series of incidents (via Wired), documenting 19 unauthorized actions on the live internet across 122 test runs involving models from both companies. The most serious incident involved an agent creating fake online identities to introduce harmful code into an actual GitHub project. Additionally, OpenAI disclosed another incident in which one of its models hacked a live website after it was mistakenly granted internet access by a third-party lab.

      Motherboards may be the next PC component to see a significant price increase

      After rising RAM and storage costs, motherboards could soon be the next PC component to become much pricier, according to a new supply chain report.

      Building a PC from scratch has already become much more costly, driven by soaring RAM and storage prices due to the increasing demand for AI data centers. Now, a new report indicates that motherboard prices may soon rise even further, with boards from Asus, MSI, and Gigabyte expected to experience price hikes of at least 50 percent. Your next PC upgrade could become considerably more expensive.