We build context-aware AI to detect harmful content and keep brands and platforms safe

In the era of authenticity, consumers crave genuine connections with brands, fuelling the surge in influencer marketing. However, harnessing the power of user-generated content (UGC) carries both promise and peril for brand image. To successfully navigate these challenges, specialised tools that grasp the nuances and context of UGC are essential.

The implementation of watermarking for AI-generated content poses several challenges. This Deep Dive explores the complexities of watermarking for different media types, including images, audio, video, and text. And highlights the difficulties in creating robust watermarking techniques that can withstand various attacks.

Unitary's goal is to support safer and kinder online connections. Intimate exchanges can often have more than one meaning. Playful banter between consenting users could in certain cases be misclassified as harmful or vice versa. We believe these subtleties are important as they shape the quality of our online experiences.

We attended TrustCon - a conference dedicated to trust and safety professionals from around the globe who are responsible for keeping the internet safe. It was an invaluable learning experience and here are some key takeaways.

In discussions about AI regulation, there has been relatively little of direct relevance to trust & safety. Most of the focus is on use by regulated organizations rather than misuse by platform users. However, as large platforms play increasingly significant roles in providing access to AI models, we might expect future regulatory efforts to more directly impact T&S.

As part of TrustCon2023 we had the privilege to organise a networking event titled 'Understanding the Mysteries of the Universe with Unitary.' Our CEO, Sasha Haco, gave a presentation on black holes and her experience of going from researching theoretical physics to co-founding an online safety AI startup.

Deepfake detection often feels like a vicious cycle. As methods are developed to detect deepfakes, the AI systems generating content evolve and bad actors change tactics. And it's not long before the metrics and thresholds for detecting deepfakes become outdated. In this article we explore how Goodhart's law, 'when a measure becomes a target, it ceases to be a good measure', can help us see things under a different light. 

Concerns about bias in social media content moderation have prompted the drafting of laws and the need for better explanations. Studies show users perceive automated moderation as more impartial with human oversight and that trust in AI moderation increases with transparency.