The Hal Times

Anthropic has disclosed that its Claude AI model played a role in the discovery of a novel CRISPR-like enzyme system. This marks an expansion of the company into wet-lab biology. The findings represent a departure from traditional computational tasks, as the model was utilised to navigate complex biological data to identify the enzyme. The research suggests that large language models may serve as effective tools for researchers in the laboratory.

The firm noted that humans remain in the loop throughout the process. Despite the success of the model in this specific task, Anthropic has stated that it has not granted Claude autonomy within its biology lab. The discovery is currently viewed as a proof of concept for integrating artificial intelligence into biological research workflows. The practical implications for future drug discovery or genomic editing remain to be seen.

Financial observers have suggested that such developments in the laboratory are occurring as the company approaches an anticipated public offering. Analysts are scrutinising the company to determine if these scientific advancements provide a clear competitive advantage in the crowded artificial intelligence market. Questions linger regarding the scalability of these methods and whether such discoveries can be reliably replicated outside of a controlled, proprietary environment. (The News International; analyticsindiamag.com; TechCrunch)

Meta Announces Updates To Muse AI Agent

Meta chief executive Mark Zuckerberg announced a series of updates for the company's Muse AI agent at the annual Connect event in Menlo Park. The developments include the integration of the agent into wearable glasses. Meta also debuted a small, Tamagotchi-like hardware device intended to serve as a mobile home for the agent.

In addition to the software updates, the company introduced camera-free AI glasses. These are designed to be lighter than previous versions and feature a battery life of up to 12 hours. The strategy marks a significant commitment to incorporating artificial intelligence into consumer hardware products. (TechCrunch; TechCrunch; TechCrunch)

From the Laboratories

•  Google DeepMind has introduced secure, server-side memory for its Private AI Compute services. (Google DeepMind)

•  NVIDIA has released documentation on using NVIDIA Warp and MjWarp to improve robotics simulation and learning workflows. (Hugging Face)

•  OpenAI has marked two years of its OpenAI Academy initiative focused on community skill development. (OpenAI)

•  Google DeepMind has released Gemini 3.8 text-to-speech capabilities. (Google DeepMind)

•  OpenAI is extending access to its Daybreak program to assist the Government of Ukraine with civilian cyber defence. (OpenAI)

•  Sam Altman discussed international cooperation and safety at the United Nations Security Council. (OpenAI)

•  Financial analysts are weighing the timing of the Anthropic IPO against broader market and safety concerns. (Reuters)

Industry Leaders Confer

AI Executives Call For Global Regulation At United Nations

The chief executive officers of OpenAI, Anthropic, and Hugging Face have formally petitioned the United Nations to pursue global regulation of artificial intelligence. In a recent session at the Security Council, the representatives argued that current oversight remains inadequate for the potential risks posed by advanced systems. The executives acknowledged that tech companies do not possess all the necessary answers regarding the policy frameworks required to govern these technologies safely.

The call for international cooperation coincides with a period of heightened debate regarding the direction of AI governance. While some factions within the United States, including former President Trump, have expressed opposition to global regulatory frameworks, industry leaders maintain that the rapid development of large language models necessitates a unified, global approach. They suggest that the risks inherent in these systems, which some describe as a threat to all humanity, cannot be mitigated through domestic policy alone.

The path forward remains uncertain as governments weigh the balance between innovation and public safety. Whether member states will coalesce around a binding international treaty or prefer voluntary guidelines remains a question for diplomats. The industry leaders argue that without a consistent global standard, the development of these systems will continue to outpace the existing capacity for effective oversight by national regulators. (Modern Ghana; Al Jazeera)

OpenAI Files Response In xAI Antitrust Lawsuit

OpenAI has moved to dismiss elements of an antitrust lawsuit brought by xAI. The company claims that recent disclosures filed with the Securities and Exchange Commission undermine the allegations put forward by its competitor. OpenAI argues that the facts provided to regulators demonstrate that its business practices remain within the bounds of market competition laws.

The legal dispute highlights the intensifying rivalry between major AI laboratories. As these firms navigate both federal scrutiny and private litigation, the court must now determine if the claims of market manipulation hold merit or if they represent standard competitive behaviour in an emerging technological sector. (The Economic Times)

In Brief

•  The Chief Justice of Victoria has announced a new strategic vision regarding the integration and management of AI technologies within the state court system. (Law360)

•  Growing public anxiety regarding data privacy and surveillance is reportedly hampering the adoption of new consumer AI devices. (UA.NEWS)

•  An Australian government agency abandoned an AI fraud detection system after the tool generated an unmanageable volume of false positives. (The Canberra Times)

New Model Release

Anthropic And OpenAI Launch Rapid Model Updates

Anthropic has released Claude Opus 5.5, the latest iteration in its suite of large language models. The release occurred shortly before OpenAI introduced two additions to its own lineup, GPT-6 Sol and GPT-6 Luna. These new offerings from OpenAI arrive at half the price of their predecessors, the GPT-5.6 equivalents. The rapid succession of these releases underscores a period of intense competition within the industry.

Analysts are currently examining the performance capabilities of these new models. Early impressions suggest that users may require further time to determine the practical differences between these versions. The shift in pricing structures indicates that vendors are prioritising affordability alongside incremental model improvements for the development community.

These releases follow a busy period of activity that included the debut of Grok 4.7 and MiMo v2.6 Flash and Pro. The speed at which these models reach the market suggests that the pace of development in large language models shows no sign of slowing. Whether these price reductions will fundamentally alter developer adoption habits remains to be seen as benchmarks continue to fluctuate. (Simon Willison)

Graphify Aims To Improve Codebase Navigation

Graphify, an open-source tool launched in April 2026, seeks to assist artificial intelligence coding assistants by converting codebases into queryable knowledge graphs. The tool aims to solve the difficulty of multi-file reasoning, a common hurdle for agents tasked with complex programming duties. Recent updates to the software focus on enhancing parser features and cross-file resolutions.

While community feedback notes a promising architecture, some developers report challenges integrating the tool into existing daily workflows. As software engineering becomes increasingly agentic, tools like Graphify represent an attempt to provide the necessary context for models to operate effectively across large, fragmented repositories. (InfoQ)

Google Releases New Gemini Text To Speech Models

Google has released two text-to-speech models, gemini-3.8-flash-tts and gemini-3.8-flash-lite-tts. These additions come with a library of over 2,000 voices. The models allow for the creation of custom voices using a 30-second audio sample from the user or from a source for which the user holds the appropriate rights. The release is accompanied by a playground interface, which allows developers to test these voices through an API implementation using a bring-your-own-key authentication method. (Simon Willison)

In Brief

•  GitHub has rebuilt the diff surface in the GitHub Copilot app to handle pull requests containing up to a million lines of code. (GitHub)

•  Qualcomm confirmed that Linux support is coming to the Snapdragon X2 Series of processors. (Hacker News)

•  A new guide from Hugging Face details how to use NVIDIA Warp and MjWarp for accelerating robotics simulation workflows. (Hugging Face)

Meta Expands Muse

Meta Introduces Muse Charm And Expands Agent Features

Meta showcased a new dedicated hardware device during the Meta Connect presentation. Known as Muse Charm, the gadget functions as a standalone agent for the Muse AI system. It features a large screen and resembles a chunky smartwatch without a wrist strap, utilising a lanyard for portability. Chief Executive Mark Zuckerberg demonstrated the device at the conclusion of the event.

Beyond the hardware, Meta announced that the Muse agent will soon appear on its smart glasses. Users may activate the system by voice command to assist with tasks such as workout guidance and data logging. These capabilities arrive as the company continues to iterate on the Muse platform.

Meta also confirmed that Muse agents are gaining individual email addresses to facilitate task completion. Furthermore, the company is enabling video chat features for users to communicate directly with their agents. These updates signal a broader effort to embed AI agents into daily wearables and routine digital tasks. The full scope of public adoption remains to be seen as these devices enter the marketplace. (The Verge; The Verge; The Verge)

Adobe Finalises Acquisition Of Topaz Labs

Adobe has completed its purchase of Topaz Labs, a software firm recognised for AI-based upscaling technology for images and video. The agreement, first announced in June, makes Topaz Labs a wholly owned subsidiary. Adobe confirmed that Topaz Labs will continue to operate as its own distinct brand following the integration.

This move brings specialised upscaling tools under the Adobe corporate umbrella, which may influence future developments in creative software workflows. While Adobe assumes control of the technology, the long-term impact on the standalone product suite remains a subject of industry observation. (PetaPixel)

Level 5 Comments On Generative AI Usage

Game developer Level 5 has addressed the use of generative AI in a recent presentation. The company provided insight into its stance on incorporating these tools into its creative process. This follows a period of increasing scrutiny regarding AI deployment within the interactive media industry.

Specifics on how these tools influence their current production pipeline were discussed for stakeholders. The extent to which these systems alter the labour practices or design philosophy of the studio remains a point of interest for the broader gaming community and its various trades. (Anime News Network)

In Brief

•  Stanford University acknowledged a violation of its own policies after using AI to edit a student photograph, altering the race and gender of the subjects. (PetaPixel)

•  A new Microsoft patent suggests the company is exploring technology that would allow video games to rewrite their own content using AI. (CNET)

•  YouTube plans to introduce custom feeds and expanded AI features for its platform later this year. (Ars Technica)