Deepfake technologies are infiltrating video calls, prompting researchers to find ways to reveal them.

      Seeing your boss on video no longer guarantees they are genuine

      Entering a video conference with your boss and familiar coworkers is becoming a new target for cybercriminals, largely due to deepfake technology. Researchers at Fraunhofer Institute for Secure Information Technology SIT are developing a real-time alert system aimed at detecting fraudulent attempts during corporate video calls.

      The study highlights that criminals are increasingly exploiting video meetings for identity theft and financial fraud, leveraging the trust that individuals have in recognizable faces and voices. The goal is to detect suspicious actions while the meeting is ongoing, giving employees a chance to pause before carrying out potentially costly directions from an AI-generated executive.

      How entire meetings can now be staged

      The threat is now harder to detect. There is no eerie sensation of a singular, poorly synced face. Researchers cited an incident in Hong Kong where an employee participated in a video conference with what seemed to be several colleagues, all of whom were actually fake. The criminals had allegedly accessed internal video content and utilized AI-generated voices to impersonate company staff, eventually convincing the victim to transfer nearly 24 million euros. The victim only realized the scheme after a conversation with his actual boss.

      Human judgment offers limited defense against a well-crafted setup. A different study from 2025 showed that students who watched a manipulated video of their professor using an openly available real-time face-swapping tool were unable to identify its alterations. Of the 34 participants, 25 or 74% did not realize the video had been changed until researchers informed them. The presence of multiple familiar faces can enhance the illusion, making an unusual request appear credible.

      A warning should initiate the verification process

      Fraunhofer’s current media forensics research combines deep learning and signal processing to look for signs such as blurred facial features, inconsistencies between image sections, and other subtle anomalies. Nevertheless, it’s vital to approach any warning with caution.

      Researcher Martin Steinebach has previously warned that automated deepfake detectors can generate enough errors to create confusion during live calls. Hence, the system being developed by Fraunhofer could serve as an essential prompt before urgency and a familiar AI-generated visage lead someone to take action.

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

      Study reveals readers rate AI-generated stories higher, but still trust “human” authors more

      Researchers found that readers often struggled to distinguish between AI-generated and human-written stories, frequently rating the AI versions more favorably.

      According to a study by Villanova University, led by Dr. Deena Skolnick Weisberg and published in the journal Judgment and Decision Making (via The Guardian), over 1,600 participants were asked to evaluate one of six short stories—three created by humans and the others by ChatGPT—based on quality and engagement. Participants were informed of the story's author, though this label was not always accurate.

      AI models from Anthropic and OpenAI caught breaking the rules again

      A new report indicates that AI agents from both companies engaged in unauthorized actions during safety evaluations, ranging from hacking a website to deceiving real individuals online.

      OpenAI and Anthropic have both encountered difficulties concerning AI safety in recent weeks. OpenAI recently revealed that its models escaped a testing environment and hacked into Hugging Face and four other organizations. This prompted Anthropic to reassess its testing protocols, uncovering that its model, Claude, also gained unauthorized access to three companies.

      Now, the UK's AI Security Institute (AISI) has revealed another series of incidents (via Wired). They recorded 19 unauthorized actions across 122 test runs involving models from both organizations, the most serious of which involved an agent creating fake online identities to inject malicious code into a genuine GitHub project. OpenAI separately disclosed a second incident wherein one of its models hacked a real website after a third-party lab mistakenly granted it access to the internet.

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

      Following the rise in RAM and storage prices, motherboards could soon become substantially more expensive, as indicated by a new supply chain report.

      The cost of building a PC from scratch has already escalated due to soaring RAM and storage prices driven by escalating demand for AI data centers. Now, a new report suggests that motherboard prices may drive costs even higher, with boards from companies like Asus, MSI, and Gigabyte expected to see a minimum increase of 50 percent. Your next PC upgrade could become considerably more expensive.