The Hal Times

Akamai has agreed to a $11.6 billion deal with Anthropic to build a cloud infrastructure for AI models. The deal, announced on September 25, will see Akamai provide cloud services to Anthropic, including content delivery and edge computing. The partnership is expected to boost Akamai's existing AI-driven tailwind, while also presenting risks to its cloud horizon.

The deal comes as part of a transformative week for US stocks, with TD SYNNEX's Q3 earnings plunging by 9.87%. Akamai's investment in Anthropic is seen as a strategic move to capitalise on the growing demand for AI infrastructure. The deal is expected to create new opportunities for both companies, with Akamai gaining a foothold in the AI market and Anthropic benefiting from a robust cloud infrastructure.

The deal is subject to regulatory approval, and both companies are expected to announce further details in the coming weeks. The partnership is seen as a significant step towards integrating AI with cloud services, and could set a new standard for the industry. However, the deal also raises questions about the long-term viability of Akamai's cloud services in the face of increasing competition. (HDFC Sky; Seeking Alpha)

US Government Upholds Pentagon Blacklist Order Against Anthropic

The US government has upheld a Pentagon blacklist order against Anthropic, following a ruling by the DC appeals court. The order, which was issued in July, restricts the company's access to certain Pentagon-funded research projects. The ruling is seen as a significant setback for Anthropic, which has been working to expand its AI capabilities.

The Pentagon's blacklist order is part of a broader effort to regulate the use of AI in defence-related projects. The order is expected to have a significant impact on Anthropic's ability to collaborate with government agencies. The company has not yet commented on the ruling, but analysts suggest that the order could lead to further restrictions on AI research in the US. (Business Today)

Crusoe Abandons $1.25 Billion Plan for Boom Turbines at AI Data Centers

Crusoe has abandoned its $1.25 billion plan to use Boom Supersonic's turbines at AI data centers. The company's new stationary power plants are no longer in Crusoe's near-term plans, according to Boom Supersonic CEO Blake Scholl. The decision is seen as a significant setback for Crusoe, which had been working to expand its renewable energy capabilities.

The abandonment of the plan is part of a broader shift in Crusoe's strategy, which has been focusing on developing new power solutions for AI data centers. The company has not yet announced a replacement for the Boom turbines, but analysts suggest that the decision could lead to further delays in Crusoe's expansion plans. (TechCrunch)

From the Laboratories

•  Meta opens early access program for new Muse features, requiring users to request inclusion. (TechCrunch)

•  Unsecured OpenAI agents posted 53 user images on the internet without the lab's knowledge. (TechCrunch)

•  The Pentagon seeks $30.3 million to build an AI-powered lie detector over five years. (MIT Tech Review)

•  Google DeepMind introduces Gemini 3.8 Live with Live Avatar. (Google DeepMind)

•  Hugging Face accelerates vision-language models with LFM2.5-VL-DSpark. (Hugging Face)

Pentagon Risk Classification

Court Upholds Pentagon Blacklist Order Against Anthropic

The ruling follows a legal challenge by the artificial intelligence firm regarding its inclusion on a list of companies identified as potential threats to the military supply chain. The court found that the Department of Defense acted within its discretionary powers when categorising the company. This decision confirms the government's authority to restrict corporate entities from certain state contracts based on perceived security vulnerabilities within their operational structures.

The dispute highlights growing tensions between the executive branch and private developers regarding the integration of large language models into national security infrastructure. While the court has upheld the designation, the precise criteria the Pentagon employed to measure this supply chain risk remain largely shielded from public scrutiny. It is unclear what specific architectural or corporate dependencies triggered the department's assessment, though the judgment establishes a precedent for how federal agencies may classify emerging technology firms.

For now, the decision reinforces the Pentagon's role as a primary arbiter of technological safety. Anthropic remains restricted from participating in specific procurement channels, a move that may influence how other federal departments approach their own vendor oversight. The court provided no immediate directive for the government to remove the designation, leaving the firm in a position of sustained exclusion from relevant defence contracts. (The Washington Post; Business Today)

Pentagon Seeks Funding For Artificial Intelligence Lie Detector

The Department of Defense has requested 30.3 million dollars to develop an improved polygraph system over the next five years. The project, currently referred to as Polygraph+ or Polygraph Next, seeks to move beyond traditional methods by incorporating artificial intelligence and machine learning scoring algorithms. The department aims to refine how biometric data is processed during security investigations.

A central component of the proposed system involves standoff sensing, a technique intended to monitor subjects from a distance. While the department has outlined its budget request, details regarding the accuracy of these machine learning models against human physiological variability remain undisclosed. It is currently unknown how the government plans to address potential biases inherent in using AI to interpret human stress responses. (MIT Tech Review)

Republican Factions Debate Future Of Artificial Intelligence Regulation

Internal divisions have emerged among Republicans regarding the pace and necessity of federal artificial intelligence regulation. While some leaders have expressed a desire to minimise oversight to encourage industry growth, conservative dissenters are pushing back against this position. These legislators suggest that a lack of clear standards could leave the nation vulnerable to unforeseen risks.

The rift highlights the difficulty of balancing economic deregulation with safety concerns within the party. Observers note that this disagreement may stall upcoming legislative efforts, as the debate exposes differing views on the role of government in managing emerging technologies. The ultimate impact on future policy remains difficult to forecast. (디지털투데이)

In Brief

•  Joe Lonsdale has warned that current regulatory pushes for artificial intelligence may inadvertently create a dangerous oligopoly. (Mena FN)

•  New Mexico Attorney General Raúl Torrez is calling for federal regulation of artificial intelligence following a security breach at the University of New Mexico. (Santa Fe New Mexican)

•  Legal experts are debating whether Australia might become a safe haven for copyright infringement by artificial intelligence developers. (insidestory.org.au)

•  The Electronic Frontier Foundation reports that DraftKings is utilising artificial intelligence to refine online behavioural advertising for sports betting. (EFF)

Building Agents

GitHub Extends Copilot With Custom Workflows

GitHub has introduced a new capability for its Copilot application aimed at developers who wish to build custom workflows. By describing a desired interface in plain English, a user may prompt an agent to generate a live surface. This canvas is designed for both the agent and the developer to share, allowing for iterative updates. The goal is to reduce the time spent adjusting to various development tools by creating a tailored environment that evolves alongside the project.

The approach reflects a broader trend toward more interactive agentic systems in software engineering. While these tools promise efficiency, they introduce new demands on the user. Engineers must now maintain strict discipline to manage these automated processes effectively. As the systems become more capable, the complexity of the underlying workflows often grows in tandem, requiring a higher degree of oversight from the human developer.

Practical experience suggests that although these agents unlock significant potential, they do not necessarily simplify the task of software engineering. Some observers argue that these tools may even increase the difficulty of the profession. Unlocking full performance requires deep technical knowledge and careful management of the agentic harness. Developers must weigh the convenience of these new canvases against the persistent need for manual control and rigorous evaluation of the resulting code. (GitHub Blog; Simon Willison)

Methods For Constructing Agent Harnesses

A recent technical overview examines the architecture of the agent harness, a critical component for managing operational layers in AI applications. The guide details two approaches to building a finance assistant: one using AWS AgentCore Harness and another utilizing LangChain with Envoy AI Gateway. These implementations provide a framework for handling memory, tool access, and cost control.

The comparison highlights significant trade-offs regarding portability and engineering effort. Choosing between these systems requires developers to consider their specific needs for observability and long-term operational ownership. Proper harness design remains essential for moving agentic systems from experimentation to reliable production environments. (InfoQ)

In Brief

•  The U.S. appeals court has upheld a ruling designating Anthropic as a supply chain risk, according to CNBC. (CNBC)

•  EasyStack has announced EAF, described as an AI-native cloud foundation for enterprise deployments. (ANTARA News)

•  FTC chair suggests that AI developers should be held liable for the conduct of their agents. (Reuters)

Music Industry Conflict

Sony And Universal Music Group Sue Suno Again

Sony and Universal Music Group have filed a fresh lawsuit against Suno. The labels claim the company's new v6 model relies upon copyright-infringing foundations. They argue that the model was trained on outputs from earlier versions of the software, which were themselves built upon music ripped from YouTube and other platforms without permission.

The legal dispute highlights a widening rift between major music publishers and developers of generative AI. Sony and Universal remain notable holdouts, having declined to sign licensing agreements with AI music firms. The labels maintain that the unauthorized ingestion of their catalogues constitutes a fundamental violation of intellectual property rights.

Suno has not yet commented on the specific allegations regarding the v6 architecture. The outcome of this case may establish a significant precedent for how AI companies source training data. For now, the music industry remains split between those seeking partnerships and those pursuing litigation to protect their established catalogues. (The Verge)

Disney Initiates Search For Director Of AI Enablement

Disney is seeking a Director of AI Enablement following an internal memo regarding companywide layoffs. A legal executive at the firm indicated that management intends to overhaul established workflows. The memo noted a necessity to take a dispassionate look at every aspect of operations to ensure greater efficiency and lower costs.

The move signals a broader transition within the studio as it integrates generative tools into its creative pipeline. While the company pursues these cost-saving measures, the impact on staff remains a primary concern for the entertainment unions. Disney has provided little detail on which specific creative departments face the most significant reorganisation under this new mandate. (Hollywood Reporter)

Bristol Venues Oppose Generative AI Artwork

Live venues in Bristol have begun to strike back against the use of generative AI artwork for promotional materials. The campaign reflects a growing trend among local arts communities concerned about the displacement of human illustrators and designers. Organisers argue that the adoption of synthetic imagery undermines the cultural value of the city's artistic scene.

The initiative puts pressure on promoters to commit to using work created by human hands. Similar concerns have surfaced at the Madison Museum of Contemporary Art, which recently debuted an exhibition featuring generative AI works despite public criticism. Both incidents underscore the rising friction between technological adoption in galleries and local calls for artistic protection. (BBC; WPR)

In Brief

•  Adobe has integrated Lightroom and Photoshop tools directly into the Google Gemini platform. (PetaPixel)

•  A man who allegedly generated eight million dollars using AI songs and bot streams has requested no prison time, claiming no perceptible harm to artists. (Rockum)