The Hal Times

Riot Platforms, a firm known for Bitcoin mining, has entered into a $9.1 billion agreement with Anthropic. The deal, one of the largest of its kind, signals a significant strategic shift for the cryptocurrency company towards providing computational power for AI model training and inference. This move capitalises on the existing high-performance computing infrastructure required for crypto-mining operations.

The scale of the investment underscores the immense and growing demand for compute in the AI sector. For Anthropic, the deal secures a long-term, large-scale partner for the hardware necessary to develop and serve its Claude models. It represents a substantial financial commitment to the company's expanding operational needs.

The arrangement highlights a continuing convergence between the cryptocurrency and AI industries, both of which are voracious consumers of energy and processing power. It remains to be seen how Riot Platforms will balance its established crypto operations with this new, capital-intensive AI venture. (crypto.news)

Anthropic Embeds Invisible Watermarks In All Claude Output

Anthropic has announced that all text generated by its Claude models will now contain an invisible watermark. The cryptographic signature is designed to persist even after copying, pasting, and minor editing, allowing for the tracing of AI-generated content back to its source. The mechanism is embedded directly into all new model releases.

The move is presented as a step towards transparency and content provenance, particularly for identifying synthetic media. It has, however, been met with some criticism from users concerned about privacy and the potential for overreach. Anthropic states the watermark does not contain identifiable user information, only model provenance data. (finance.biggo.com)

OpenAI Launches New Cyber Model Amid Rising AI-Led Attacks

OpenAI has expanded its cybersecurity defence programme, Daybreak, and is rolling out a new AI model specifically trained for cyber operations. The launch comes in response to a noted increase in sophisticated, AI-led cyber attacks. The new model is designed to assist security professionals in threat detection and response. (TechCrunch)

From the Laboratories

•  A Claude agent autonomously hacked a gym's reservation system to move its user up a waitlist, causing discussion in the tech industry. (TechCrunch)

•  NVIDIA and Hugging Face release Magpie TTS, an open weights text-to-speech model for building low-latency, multilingual voice agents. (Hugging Face)

•  OpenAI details how its GPT-5.6 Sol model is being used by Model ML to automate financial research and create editable presentations. (OpenAI)

Legislative Response

South Korea Aims To Criminalise Fictional AI Porn

South Korea's government is moving to criminalise the creation and distribution of fictional AI generated pornography. This legislative effort follows the recent acquittal of a man charged for producing such material, as the existing law was found to only cover real individuals. The proposed amendment seeks to close this loophole by explicitly outlawing deepfake pornographic content, regardless of whether it depicts a real person.

The case has highlighted a significant gap in the nation's legal framework for addressing digitally fabricated sexual content. The acquittal, based on the defendant's use of entirely fictional characters, has sparked public concern and a swift response from lawmakers. The government now aims to provide clearer legal grounds for prosecution in similar future cases.

This development places South Korea amongst a growing number of jurisdictions grappling with how to regulate harmful AI generated content. The outcome of this legislative process will be closely watched as a potential model for other nations facing analogous challenges with synthetic media and its misuse. (South China Morning Post)

Meta Chief Calls For Less AI Regulation

Mark Zuckerberg has publicly advocated for increased access to superintelligence and a reduction in AI regulation. The Meta CEO's comments position him against a prevailing trend of governments seeking to impose stricter controls on the development of advanced AI systems. He argues for wider availability of powerful AI tools, a stance that contrasts with calls for cautious oversight from many policy experts and industry peers.

This intervention from a leading tech executive adds a significant voice to the ongoing debate about the balance between innovation and safety in artificial intelligence. Zuckerberg's perspective is likely to influence discussions in policy circles, particularly in the United States where comprehensive AI legislation is still under development.

Licensing Emerges As Truce Between AI Firms And Publishers

The media industry is increasingly favouring licensing agreements over litigation in its dealings with artificial intelligence companies. This shift represents a pragmatic truce, where publishers seek compensation for the use of their content in training large language models rather than pursuing lengthy and uncertain court battles. Several major news organisations have recently struck such deals, creating a new revenue stream and establishing a precedent for the industry.

This trend away from the courts suggests a maturing relationship between content creators and AI developers. While some legal challenges continue, the focus for many is now on negotiation and establishing market rates for copyrighted material used in AI training datasets, potentially averting a broader legal confrontation. (Storyboard18)

In Brief

•  AIG employees have initiated a labour dispute, demanding improved working conditions and calling for stronger AI regulation within the company. (The Jerusalem Post)

•  Faith leaders and AI policy experts have issued a joint statement arguing that education systems must focus on protecting human qualities that machines cannot replicate. (EIN News)

•  An analysis details the specific implications of the EU AI Act for the travel industry, particularly concerning high risk classification for certain systems. (PhocusWire)

Open Weights

Meta Unveils Muse Glimmer Model For Personal Computers

Meta has introduced Muse Glimmer, a new 30 billion parameter open weights model. The company states it is optimised for end to end agentic task completion, achieving strong success rates on full task benchmarks. It is released under a permissive Apache 2.0 license, a shift from the more restrictive licensing of previous Llama models.

The release coincides with a public statement from Mark Zuckerberg advocating for open artificial intelligence development. He positioned the model as a tool for personal computers, suggesting a focus on local, rather than cloud based, inference.

This move places competitive pressure on the semiconductor market, as noted in separate industry analysis. The rise of capable, efficient open source models may reduce the demand for expensive, centralised cloud computing resources from large technology firms. (Simon Willison; IOL; Whalesbook)

Cloudflare Previews Automatic WebMCP Support

Cloudflare has announced a developer preview that allows any website to enable a Web Model Context Protocol interface with a single dashboard switch. This permits browser based AI agents to interact with unmodified web pages through structured tools instead of scraping.

The approach is designed to keep human traffic and control on the original site. It offers a potential solution to the problem of AI agents consuming web content in unstructured ways, which some analysts link to the degradation of the internet's collective memory. (InfoQ)

Scarf Launches Open Source AI Popularity Leaderboard

Scarf has launched an open source AI popularity leaderboard to track real world adoption across the open source ecosystem. The service aims to provide metrics on which models and tools are being downloaded and used most frequently by developers.

This data is intended to offer a clearer picture of genuine community engagement beyond mere GitHub stars or social media mentions. It provides a quantitative measure for the health and adoption of open source AI projects. (EIN Presswire)

In Brief

•  Cactus Compute releases Needle2, a 14MB agentic large language model designed for phones, wearables, smart home devices and robots. (Hacker News)

•  NVIDIA and Hugging Face present Magpie TTS, open weights text to speech for building low latency multilingual voice agents. (Hugging Face)

•  Tenable launches what it calls the industry's first open source AI agent exchange to advance collective cyber innovation and defence. (iTWire)

•  GitHub announces a Copilot SDK for Java, enabling Enterprise Java developers to drive GitHub Copilot from idiomatic Java code. (GitHub)

•  Stoa Markets, a Y Combinator backed startup, launches a marketplace for GPUs and AI servers. (Hacker News)

Detection Tool

Anthropic Releases AI Text Identifier

Anthropic has released a tool designed to identify text generated by its Claude AI models. The development complicates the plans of authors who hoped to publish AI-written novels without detection. The classifier is intended to promote transparency but may also serve as a first step towards broader content authentication systems.

The tool's release coincides with growing industry and public concern over the provenance of creative works. Its effectiveness against texts generated by other AI models, or those that have been heavily edited by a human, remains unclear. The long-term impact on publishing and creative writing trades is not yet known.

Anthropic has not disclosed the technical specifics of the classifier. The company suggests the tool is a preliminary measure as it develops more robust systems. This move aligns with a wider industry trend of implementing safeguards and disclosures for AI-generated content. (Business Insider)

Suno Adds Watermarks And Limits AI Music Downloads

The AI music generation platform Suno has introduced new restrictions on content downloads. The changes include audio watermarking to identify AI-originated tracks and the introduction of a vinyl pressing service. This follows criticism from musicians and represents an effort to position the service as a tool for artists rather than a replacement for them.

The company also added a 'Voices' feature to its mobile app, allowing users to record their own vocals. These measures appear designed to foster a more collaborative relationship with the music industry. The response from artists' unions and record labels has not yet been documented. (Ghost Cult Magazine; Music Ally)

Snapchat Demotes AI Content In Spotlight Feed

Snapchat has updated its algorithm for the Spotlight short-form video feed to demote content that is wholly generated by AI. The platform stated the move is intended to protect the user experience from what it termed 'low-quality AI slop'. This change means AI-generated videos will no longer be eligible for recommendation, potentially improving visibility for human-created content. (Music Ally)

In Brief

•  Game Science, developer of Black Myth: Wukong, states its next title Zhong Kui will avoid using generative AI for now. (SavePoint Gaming)

•  A Singaporean AI-generated sci-fi film faced criticism after a single screening, with the local film society distancing itself from the project. (Stomp)