The Hal Times

In its latest financial filing, Anthropic reported revenue of $16.8 bn, a 46% increase year‑on‑year, and said it has reached profitability ahead of a planned initial public offering. The figures were disclosed in a Bloomberg report and confirm the company’s rapid expansion since its 2023 launch. The firm attributes the growth to heightened demand for its Claude series of large language models and a growing enterprise client base. The announcement arrives as investors watch a crowded AI market where rivals such as OpenAI and Google continue to raise capital.

The surge lifts Anthropic’s valuation prospects and gives it leverage in upcoming fundraising rounds. Analysts note that the $16.8 bn turnover places the firm among the few AI‑focused firms to post double‑digit growth without a public listing. The profitability claim, also made by Bloomberg, suggests that Anthropic’s subscription‑based API model is now covering its heavy compute and research costs. This could persuade hesitant investors to back the forthcoming IPO, potentially adding a new heavyweight to the NASDAQ AI roster.

What is new is the combination of a near‑term profit claim with a sizeable revenue jump, both rare in a sector still dominated by venture‑backed loss‑making players. Anthropic’s latest Claude model, Claude‑3.5, launched earlier this year, reportedly improves reasoning speed and reduces hallucinations, which may have spurred enterprise uptake. The company also announced a partnership with TSMC that helped it secure cheaper silicon, a factor mentioned in a separate Bloomberg story on TSMC’s earnings beat.

Uncertainties remain about the timing and pricing of the IPO, and whether the profit margin can be sustained as competition intensifies. The firm has not disclosed the exact profit figure, leaving analysts to guess at margin depth. Moreover, regulatory scrutiny of AI systems could affect future revenue streams, especially as the company prepares to add official US voting information to its Claude model ahead of the midterms, a move that may attract both users and critics. (Bloomberg.com)

Anthropic Adds Official US Voting Information to Claude

Anthropic will integrate official United States voting data into its Claude assistant ahead of the November midterm elections, according to The News International. The update will allow users to query poll locations, ballot measures and candidate details directly within the chatbot, aiming to improve civic engagement and reduce misinformation. The move follows pressure on AI firms to provide reliable public‑service content as political discourse increasingly migrates to generative platforms.

The addition raises questions about data freshness and liability, as Anthropic must keep the information current across thousands of jurisdictions. Critics warn that embedding electoral data could expose the model to manipulation attempts, while supporters argue that a trusted AI conduit could counteract false narratives circulating online. Anthropic has not disclosed how it will source or verify the voting information, leaving the effectiveness of the feature open to scrutiny. (The News International)

Figma Announces New Features to Define How Agents Work

Figma announced a suite of new tools designed to define how autonomous agents interact with its design platform, reported by TechCrunch. The features include a visual scripting interface, permission controls, and a sandboxed execution environment that lets developers embed AI‑driven assistants into design workflows without compromising file integrity. Figma also launched its code layers capability in open beta, enabling agents to generate and edit front‑end code directly from design components.

The enhancements aim to standardise agent behaviour across the rapidly expanding ecosystem of AI‑powered design assistants. By exposing a clear API and safety guards, Figma hopes to attract third‑party developers while reassuring enterprise customers of data security. However, the company acknowledges that agents may still produce unexpected outputs, and the long‑term impact on traditional design roles remains to be seen. (TechCrunch)

Ricursive Intelligence Bets on AI‑Designed Chips

Ricursive Intelligence, a startup profiled by TechCrunch, is developing AI systems that design their own semiconductor layouts to alleviate the industry’s compute shortage. The founders describe a closed loop where a generative model proposes a chip architecture, a simulator evaluates performance, and the results feed back into the model for the next iteration. This automation aims to compress design cycles from months to weeks, potentially lowering costs and expanding access to custom silicon.

The approach promises to democratise high‑performance compute by reducing reliance on expensive human expertise. Yet the technology is still experimental; the company has not disclosed benchmark results or production timelines. Industry observers caution that integrating AI‑designed chips into existing fabrication flows may encounter compatibility hurdles, and regulatory scrutiny of AI‑generated hardware could slow adoption. (TechCrunch)

From the Laboratories

•  Anthropic warns of AI trade risks before the end of humanity. (Bloomberg.com)

•  TSMC earnings beat and Anthropic declares profitability. (Bloomberg.com)

•  Gudea offers AI‑driven early warning of social‑media crises. (TechCrunch)

•  Flock claims its microphones hear only gunshots, despite contrary website statements. (TechCrunch)

•  HiRO‑ACEv2 demonstrates extreme‑weather simulation at 3 km resolution for reduced cost. (Hugging Face)

•  Amp extends OpenAI agents beyond coding with GPT‑6 Astra and related models. (OpenAI)

•  SentinelOne cuts malware analysis time from weeks to hours using OpenAI Daybreak. (OpenAI)

•  Third Moment Research runs 10,000 agents for over 36 hours with GPT‑6 variants. (OpenAI)

•  Block builds AI‑powered security solutions with OpenAI technology. (OpenAI)

•  LiquidAI releases LFM2.5‑24B‑A2B for fast, private on‑device agents. (Hugging Face)

Courts Watch

States Sue Over ChatGPT Safety

Eight US attorneys general have filed a lawsuit against OpenAI, alleging that ChatGPT's responses encourage harmful behaviour. The complaint, filed in a New York court, claims the AI system 'nudges' users towards illegal or unethical actions. The states argue that OpenAI has failed to implement adequate safeguards, putting public safety at risk.

The lawsuit seeks court-ordered changes to OpenAI's operations, including a requirement to disclose how the AI system is trained and to implement stricter content moderation. The case highlights growing concerns about the safety and accountability of AI systems. OpenAI has not yet commented on the lawsuit.

The legal action comes as AI systems increasingly integrate into daily life, raising questions about their impact on public behaviour. The case may set a precedent for how regulators and courts will address AI safety in the future. The outcome remains uncertain, as the legal process unfolds. (inkl)

AI Data Center Lawsuit Expands

A New Jersey AI data centre has been sued again, with a new class action alleging environmental harm. The lawsuit, filed in a New Jersey court, claims the data centre has caused significant damage to 1,013 homes in the surrounding area. The plaintiffs argue that the centre's operations have led to increased noise, pollution, and property devaluation.

The case is part of a broader legal challenge against AI infrastructure projects. The outcome could influence how such facilities are regulated in the future. The plaintiffs seek compensation for the harm caused and a halt to the data centre's operations until safety measures are implemented.

EU AI Act Rules Drive HR Tech Acquisition

HR Path has acquired SnapAI to help clients comply with the EU AI Act. The deal reflects growing demand for AI tools that meet regulatory requirements. SnapAI's technology will assist HR Path's clients in ensuring their AI systems adhere to the EU's strict guidelines on transparency and accountability.

The acquisition highlights the commercial opportunities in AI compliance. As AI regulations tighten, companies are investing in tools to help them navigate the legal landscape. The EU AI Act, set to take effect in 2025, is expected to have a significant impact on the AI industry. (Outsource Accelerator)

In Brief

•  A family demands OpenAI shares after alleging the company used children's ChatGPT data to train AI. (Law Commentary)

•  Most New York school districts lack AI policies, raising concerns about student safety and data privacy. (NCPR: North Country Public Radio)

•  A proposal in Vietnam suggests a separate copyright law for AI-generated content, citing the speed of production. (Vietnam.vn)

•  AI labs are purchasing data from defunct startups to train their models, with one firm earning $25,000 for its codebase. (MIT Technology Review)

Anthropic Internal Tools

Anthropic Rebuilds Source Control Systems

Anthropic reports a profound increase in development velocity over the last six months. In May 2026, the company merged approximately 2,500 code changes per day. By September, that figure reached 30,000 daily merges. The existing version control infrastructure, based on Git, proved insufficient to manage this volume of activity.

The engineering team elected to design a custom source control architecture to handle the increased load. This decision highlights the limitations of traditional tools when faced with the output levels generated by modern, AI-assisted development cycles. The company documented the specific technical hurdles created by this pace of change.

The move reflects a wider trend among organizations attempting to scale software development in an era of rapid iteration. While standard tools have long served as the industry baseline, the sheer density of commits now occurring at elite AI labs appears to be pushing these foundations past their operational limits.

Mistral Releases Massive Open Weight Model

Mistral has released Le Chonk, an open weight model featuring 1 trillion parameters. This model arrives as the industry continues to explore the limits of scale for publicly accessible weights.

The release offers developers significant compute capacity for experimentation, though the hardware requirements to run such a model locally remain substantial. It marks another step in the competition to provide high-parameter models for the open ecosystem. (Startup Fortune)

Meta Secures Processing For AI Glasses

Meta has extended its Private Processing capability to its line of AI glasses. The architecture employs a combination of confidential virtual machines and trusted execution environments to protect user data.

By utilizing remote attestation and OHTTP routing, the system attempts to process persistent user context within the cloud while maintaining security. This approach seeks to balance the compute requirements of on-device AI with the privacy demands of wearable hardware. (InfoQ)

In Brief

•  Kunlun Tech has released an open-source AI game system designed to function on a single RTX 5090 graphics card. (TechNode)

•  Anthropic researchers have cautioned that small, weight-based AI models could potentially self-replicate through Internet of Things devices. (Ars Technica)

•  Liquid AI has introduced LFM2.5-24B-A2B, a model built specifically for the needs of fast on-device agents. (Liquid.ai)

•  The Allen Institute for AI has detailed HiRO-ACEv2, a model capable of simulating extreme weather events at a 3 kilometre resolution. (Hugging Face)

•  Typesafe has released Jev version 1.14.0, which includes new experimental features for developers. (Hacker News)

Transparency In Music

YouTube Music Introduces Labels For AI-Generated Content

YouTube Music has begun implementing new labelling requirements for content created with artificial intelligence. These tags are intended to inform listeners when they encounter tracks featuring AI-generated songs or vocals. The initiative serves as a response to the rapid rise of synthetic media across streaming platforms. By marking such material, the company aims to provide clarity regarding the provenance of musical works.

The rollout begins with YouTube Music but is expected to extend to the broader YouTube platform in the coming weeks. Observers note that this move mirrors a growing industry trend toward greater transparency in digital media. While the service identifies the presence of synthetic elements, the specific technical methods used to detect these files remain under discussion. Questions also persist regarding how these labels will align with diverse international regulatory requirements, including those in India.

The practical application of these labels is designed to assist users in distinguishing between human performances and those assisted by software. It marks a shift in how major streaming providers approach the integration of synthetic vocals and instrumentation. As the technology matures, such disclosures may become standard practice across all digital entertainment platforms. (YouTube Music; NewsBytes; Buzzincontent; MediaNama)

Actress Mia Goth Challenges Instagram Over Synthetic Imagery

Actress Mia Goth has publicly criticised Instagram for the platform's role in circulating an artificial image of her. The image depicted the performer wearing a t-shirt featuring a caption about her personal life. The viral nature of the post has prompted a response from Goth regarding the spread of non-consensual synthetic media on social networks.

The incident highlights ongoing difficulties in managing the proliferation of manipulated images online. While the platform continues to host a wide array of visual content, artists are increasingly seeking accountability for how their likenesses are used and distributed. (PetaPixel)

Character.AI Bankruptcy Raises User Data Concerns

The bankruptcy filing of Character.AI has triggered a multi-state investigation led by New York Attorney General Letitia James. A bipartisan coalition of 40 attorneys general has filed a notice of interest to address potential regulatory issues regarding consumer data. The regulators are seeking to ensure that personal information held by the company receives appropriate protections throughout the liquidation process.

The legal action reflects a broader focus on how AI firms manage data as they transition through insolvency. The future status of user conversations and personal records remains a subject of intense scrutiny by state officials. (The Verge)

In Brief

•  McKinsey & Company is incorporating an AI simulation of a client into its interview process to test candidates on realistic consultancy scenarios. (The Verge)

•  Eight states have initiated lawsuits against OpenAI, alleging that ChatGPT presents safety risks to children and misleads consumers. (The Verge)

•  The emergence of what is being termed grandmaslop, or AI-generated imagery featuring loved ones, is causing unease among younger social media users. (PetaPixel)