The Hal Times

A disagreement has emerged regarding the allocation of the 1.5 billion dollar copyright settlement involving Anthropic. Individual authors are currently expressing concern that book publishers and literary agents are moving to claim a disproportionate share of the funds. The tension centres on whether compensation should primarily reach those whose creative work was utilised during the training of large language models or if intermediaries deserve a larger slice of the financial pie.

The dispute highlights the ongoing complexity of valuing intellectual property in the age of generative systems. While the settlement represents a significant milestone in resolving legal challenges, the internal friction suggests that the path to distribution is far from clear. Authors argue that their original contributions form the foundation of the technology, while entities representing them often cite contractual rights to revenue streams.

What remains unknown is the precise mechanism that will be employed to resolve these claims. Both publishers and individual creators are watching the developments closely to see if a formal mediation process or a new round of litigation will be required to settle the matter. For now, the funds remain subject to these conflicting demands while legal experts assess the standard practices governing such collective settlements in the technology sector. (TechCrunch)

Anthropic Initiates Commerce Agent Programme

Anthropic has announced a new initiative to launch AI commerce agents, partnering with financial institutions Visa and Mastercard. The effort aims to integrate advanced model capabilities into the transaction ecosystem, providing automated assistance for commercial activities.

While the specific technical architecture remains proprietary, the collaboration signifies a move to embed large language models directly into payment networks. Observers note that this partnership could alter how businesses handle financial queries and transaction processing, though the practical security implications and the scale of the initial deployment remain subjects for further evaluation.

OpenAI Prepares For Financial Ratings

Both OpenAI and Anthropic are reportedly seeking investment-grade ratings as they look toward the 11.7 trillion dollar bond market following their respective post-IPO transitions. Securing these ratings is a standard procedure for companies aiming to manage large-scale capital requirements through debt instruments.

This move indicates a broader strategy to bolster balance sheets as the firms scale their infrastructure. Financial analysts suggest that the creditworthiness of AI laboratories will increasingly be judged by the reliability of their underlying models and their long-term ability to generate revenue, though the volatility inherent in this sector makes such ratings particularly difficult to forecast.

From the Laboratories

•  OpenAI is partnering with AIRPPU and WAN-IFRA to implement an AI programme designed to assist Ukrainian news organisations in strengthening their journalistic practices. (OpenAI)

•  Jakub Pachocki of OpenAI has published a reflection on the challenges of aligning increasingly capable artificial intelligence with human interests. (OpenAI)

•  Data released by OpenAI indicates that the introduction of coding agents is currently reshaping internal research workflows and increasing experiment velocity. (OpenAI)

Legal Appeal

ANI Appeals Delhi HC Ruling On OpenAI

ANI has moved to the Division Bench of the Delhi High Court to contest an earlier ruling concerning the use of its news content by OpenAI. The initial judgement, delivered by a single judge, found that the defendant's use of proprietary material for training large language models qualified as fair dealing under Indian copyright law. ANI argues that this interpretation poses a significant threat to content creators and the protection of intellectual property.

The agency is seeking to overturn the interim relief granted to the defendant. This legal skirmish highlights the tension between the development of generative AI systems and the rights of news organisations to control their output. The court must now weigh the interests of technological advancement against established copyright protections.

The outcome of this appeal may establish a precedent for how artificial intelligence companies operate within the Indian market. While the matter remains before the judiciary, the implications for data usage and compensation for media entities are substantial. It is currently unclear how long the proceedings will take to resolve. (bestmediainfo.com; storyboard18.com)

Thailand And South Korea Collaborate On Copyright

Thailand and South Korea have initiated a deeper partnership to address copyright concerns in the age of artificial intelligence. Officials are focusing on the prevention of piracy and the protection of intellectual property as machine learning models proliferate across both nations.

The cooperation aims to streamline enforcement and share expertise regarding the legal challenges posed by automated content generation. While details regarding specific legislative changes remain thin, the move marks a coordinated effort to address the risks posed by digital content scraping and unauthorised data usage in their respective regions. (nationthailand.com)

Protesters Call For Regulation In Poland

Demonstrators in Poland, using robotic displays, marched on the Digital Ministry to call for stricter oversight of artificial intelligence. The protest highlighted growing public anxiety regarding the current lack of clear regulatory frameworks governing the deployment of autonomous systems.

The organisers stated that voluntary measures are insufficient to manage the risks inherent in large-scale machine intelligence. They demand formal government intervention to ensure transparency and accountability. The Ministry has not yet issued a formal response to the specific demands presented during the march. (en.bd-pratidin.com; Startup Fortune)

In Brief

•  The Institute for AI Policy and Strategy is inviting applications for its fully-funded AI Policy Fellowship 2027. (Opportunity Desk)

•  A new briefing from the Information Technology and Innovation Foundation suggests that current AI policy should prioritise innovation as of September 8, 2026. (Information Technology and Innovation Foundation)

•  Legal analysts suggest that India's existing DABUS copyright framework may be ill-suited for the needs of everyday artificial intelligence users. (MLex)

New Security Tool

Google Opens Source Agentic Framework Mantis

Google has released Mantis, an open-source agentic framework built to automate the full lifecycle of software vulnerability management. The tool handles identifying, validating, reproducing, and fixing security flaws in code. Google developed the system to address a persistent issue in modern security automation, where conventional AI-powered code scanners frequently produce false positives or report hallucinated vulnerabilities.

Mantis operates as an agentic harness, designed to move beyond simple pattern matching. By automating the validation process, the framework intends to filter out unreliable findings before they reach a human developer. This shift is expected to save time and reduce the frustration caused by inaccurate alerts in complex codebases.

The release comes as the industry grapples with the reliability of automated systems in production environments. While Mantis provides a new mechanism for managing these tasks, its effectiveness in varied enterprise environments remains to be seen. Developers interested in the framework may now examine the underlying logic and implementation strategies provided by the project. (InfoQ)

Updates To Llms And Deployment Tools

The llm command-line tool has been updated to version 0.35, adding support for the gpt-6-astra model from OpenAI. This release allows users to interface with the new model directly through the terminal. Simultaneously, developers have been experimenting with agentic coding assistants, such as using Claude Fable 5.1 within Claude Code to generate web applications. Recent work includes the creation of a browser-based video compressor using a WebAssembly build of FFMPEG.

Separately, researchers at Columbia and Arklex AI are addressing the gap between AI agent demonstrations and reliable production deployment. At a recent presentation, Zhou Yu highlighted the use of synthetic user personas and trajectory entropy to evaluate multi-turn agents. By integrating these methods into CI/CD pipelines, teams aim to catch edge cases earlier and improve the reliability of self-learning workflows. (Simon Willison; Simon Willison; InfoQ)

Industry Groups Advance Open Source Stack

Major technology firms including Alibaba Cloud, Ant Group, Cambricon, and Huawei gathered in Shanghai at PyTorch Conference China to coordinate on the open source AI stack. The collaboration seeks to unify efforts across the industry, focusing on the infrastructure required to support open source development.

In a separate development, Mistral has announced a fundraising round of €3B. The company stated its intention to focus on the development of sovereign, open-weight AI models. These moves represent a wider industry shift toward maintaining open standards as a counterpoint to closed proprietary ecosystems. (PR Newswire; Mistral)

In Brief

•  The Shannon 3.0 AI pentester has updated its capabilities to navigate past Aikido and XBOW security measures. (TestingCatalog AI News)

Lawsuits Against OpenAI

Seattle Times And Newsday Sue OpenAI And Microsoft

The Seattle Times and Newsday have initiated legal action against OpenAI and Microsoft. The two outlets allege copyright infringement, claiming the companies utilised their journalism as training data for artificial intelligence models without permission. The lawsuit asserts that these models frequently reproduce passages from their original reporting in response to user queries.

This case mirrors similar litigation filed by other media outlets against the technology firms. The central grievance remains the unauthorised use of protected editorial content to refine machine learning systems. The publishers argue that their work is being exploited to generate outputs that may compete with the source material itself.

The legal teams must now demonstrate the extent to which the models rely on the specific datasets provided by these newsrooms. While the defendants have previously maintained that such training falls under fair use, the courts have yet to provide a definitive ruling on the matter. The outcome will likely influence how information is harvested for large language models in the future. (The Verge)

NVIDIA And Adobe Introduce RTX Spark

NVIDIA and Adobe are collaborating on a new tool titled RTX Spark, which introduces local AI agents directly into Photoshop. The partnership seeks to improve the reception of artificial intelligence among design professionals. The tool uses a half-frame trick to assist with video generation, though initial demonstrations have left some observers uncertain.

While the technology offers new capabilities for creative workflows, practitioners remain cautious. Questions persist regarding the utility of these features and their impact on the creative process. The industry is watching to see if local integration can provide a more agreeable experience than cloud-based alternatives. (Creative Bloq; Creative Bloq)

Artificial Intelligence In Film And Games

Danny Boyle has confirmed the use of generative artificial intelligence in his new film, Ink. Meanwhile, the gaming industry continues to integrate automated voice systems. NHL 27 will feature AI-generated voiceover for its sports commentators, marking a shift in how audio assets are produced for digital entertainment.

These developments highlight the increasing presence of automated tools in media production. While directors and developers experiment with these methods to streamline their output, the long-term effects on labour and authenticity remain matters of debate within the creative trades. (IGN)

In Brief

•  Design agencies are finding that AI-written briefs often fail to connect them with the right creative talent. (Creative Bloq)

•  Studio Ghibli reports that Hayao Miyazaki is not familiar with the nature of generative AI. (Creative watch)

•  Quebec courts have moved to restrict the use of generative AI by judges. (Creative watch)

•  Sonos has detailed plans for custom agents that allow listeners to create AI personas with their own voices and models. (Music Ally)

•  Recent commentary notes the prevalence of low-quality, AI-generated images in online spaces. (PetaPixel)