Thousands of abusive videos were produced last year--that researchers know of. In 2025, new data show, the volume of child pornography online was likely larger than at any other point in history. A record 312,030 reports of confirmed child pornography were investigated last year by the Internet Watch Foundation, a U.K.-based organization that works around the globe to identify and remove such material from the web. This is concerning in and of itself. It means that the overall volume of child porn detected on the internet grew by 7 percent since 2024, when the previous record had been set.

Then its sales blew up. But fails to meet the lofty goals of its own ideas. Shortly before the December 2 release of horror game, developer Santa Ragione shared some news: the game would not be available on Valve's mega platform, Steam . Valve had already banned an early, incomplete version of the game two years ago and offered, according to Santa Ragione, little clarification about why at the time. Then, hours before the game's release, the Epic Games Store banned as well.

As large language models (LLMs) become ubiquitous in our daily tasks and digital interactions, associated privacy risks are increasingly in focus.

As large language models (LLMs) become ubiquitous in our daily tasks and digital interactions, associated privacy risks are increasingly in focus.

OpenAI says the latest version of its text-to-video artificial intelligence (AI) tool Sora was downloaded over a million times in less than five days - hitting the milestone faster than ChatGPT did at launch. The app, which has topped the Apple App Store charts in the US, generates ten second long realistic-looking videos from simple text prompts. The figures were announced in an X post from Sora boss Bill Peebles, who said the surging growth came even though the app was only available to people in North America who had received an invite. The Sora app - which makes it easy for users to post videos they have created to social media - has resulted in a deluge of videos on social feeds. Some have included depictions of deceased celebrities such as musicians Michael Jackson and Tupac Shakur.

Generative AI does not only replicate human creativity but also reproduces deep-seated cultural biases, making it crucial to critically examine how concepts like ugliness are understood and expressed by these tools. This study investigates how four different generative AI models understand and express ugliness through text and image and explores the biases embedded within these representations. We extracted 13 adjectives associated with ugliness through iterative prompting of a large language model and generated 624 images across four AI models and three prompts. Demographic and socioeconomic attributes within the images were independently coded and thematically analyzed. Our findings show that AI models disproportionately associate ugliness with old white male figures, reflecting entrenched social biases as well as paradoxical biases, where efforts to avoid stereotypical depictions of marginalized groups inadvertently result in the disproportionate projection of negative attributes onto majority groups. Qualitative analysis further reveals that, despite supposed attempts to frame ugliness within social contexts, conventional physical markers such as asymmetry and aging persist as central visual motifs. These findings demonstrate that despite attempts to create more equal representations, generative AI continues to perpetuate inherited and paradoxical biases, underscoring the critical work being done to create ethical AI training paradigms and advance methodologies for more inclusive AI development.

3D scene understanding has been transformed by open-vocabulary language models that enable interaction via natural language. However, the evaluation of these representations is limited to closed-set semantics that do not capture the richness of language. This work presents OpenLex3D, a dedicated benchmark to evaluate 3D open-vocabulary scene representations. OpenLex3D provides entirely new label annotations for 23 scenes from Replica, ScanNet++, and HM3D, which capture real-world linguistic variability by introducing synonymical object categories and additional nuanced descriptions. By introducing an open-set 3D semantic segmentation task and an object retrieval task, we provide insights on feature precision, segmentation, and downstream capabilities. We evaluate various existing 3D open-vocabulary methods on OpenLex3D, showcasing failure cases, and avenues for improvement. The benchmark is publicly available at: https://openlex3d.github.io/.

An investigation by the UK advertising watchdog has found a number of shocking ads in mobile gaming apps that depict women as sexual objects, use pornographic tropes, and feature non-consensual sexual scenarios involving "violent and coercive control". The Advertising Standards Authority (ASA) used avatars, which mimic the browsing behaviour of different gender and age groups, to monitor ads served when mobile games are open and identify breaches of the UK code. While most of the thousands of promotions served to the avatars complied with UK rules, the watchdog identified and banned eight that featured "shocking" content that portrayed women in a harmful way. Two ads promoting an artificial intelligence chatbot app, Linky: Chat With Characters AI, began with a woman dressed in a manga T-shirt, a short skirt and large bunny ears dancing in a bedroom with text reading: "Tell me which bf [boyfriend] I should break up with." The ad moved on to animated content featuring text conversations with three manga-style young men.

STEM fields are traditionally male-dominated, with gender biases shaping perceptions of job accessibility. This study analyzed gender representation in STEM occupation images generated by OpenAI DALL-E 3 \& Black Forest FLUX.1 using 150 prompts in three linguistic forms: German generic masculine, German pair form, and English. As control, 20 pictures of social occupations were generated as well. Results revealed significant male bias across all forms, with the German pair form showing reduced bias but still overrepresenting men for the STEM-Group and mixed results for the Group of Social Occupations. These findings highlight generative AI's role in reinforcing societal biases, emphasizing the need for further discussion on diversity (in AI). Further aspects analyzed are age-distribution and ethnic diversity.

As part of a collaboration between Better Images of AI and Cambridge University's Diversity Fund, Hanna Barakat was commissioned to create a digital collage series to depict diverse images about the learning and education of AI at Cambridge. Hanna's series of images complement our competition that we opened up to the public at the end of last year which invited submissions for better images of AI from the wider community – you can see the winning entries here. Hanna provides her thoughts on the challenges of creating images that communicate about AI histories and the inherent contradictions that arise when engaging in this work. As outlined by the Better Images of AI project, normative depictions of AI continue to perpetuate negative gender and racial stereotypes about the creators, users, and beneficiaries of AI. The lack of diversity--and the problematic interpretation of diversity--in AI-generated images is not merely an'output' issue that can be easily fixed.