Edition, Wed, 23 Sep 2026.

China probes DeepSeek and Moonshot over alleged data leaks to Anthropic

China’s internet regulator launched an investigation into two major Chinese AI labs after Anthropic accused them of routing millions of user queries through Claude at unauthorized scale.

China’s internet regulator launched an investigation into two major Chinese AI labs after Anthropic accused them of routing millions of user queries through Claude at unauthorized scale.

According to reports by The Information and NDTV, the Cyberspace Administration of China summoned seven companies named in Anthropic's September 2026 threat report and narrowed its focus to DeepSeek and Moonshot AI. Anthropic alleged that DeepSeek routed more than 12.1 million exchanges to Claude over a 14-day period in July, with some users believing they were interacting with DeepSeek's own models. Moonshot AI was accused of routing nearly 300,000 requests to Claude over 10 days through thousands of fraudulent accounts, using Claude responses to train its Kimi model family.

Anthropic characterized the activity as unauthorized access and capability harvesting at industrial scale, pointing to routed requests containing surveillance data, internal corporate credentials, and other sensitive information. The regulator's probe is notable because it centers on the transfer of user prompts and potential data-privacy violations, not solely on intellectual-property questions around model output distillation.

Snorkel AI triples its valuation to $3.5 billion in a massive Series E

The startup that sells AI training datasets and reinforcement learning environments has seen its revenue grow 18-fold in a year, riding an insatiable appetite from AI labs for high-end training data.

The startup that sells AI training datasets and reinforcement learning environments has seen its revenue grow 18-fold in a year, riding an insatiable appetite from AI labs for high-end training data.

Led by Insight Partners and S32, the $350 million Series E valued the seven-year-old company at nearly triple the $1.3 billion it commanded in a Series D 17 months ago. Co-founded by Stanford AI researcher Alex Ratner, Snorkel shifted last year from data-labeling software to a data-as-a-service model that combines synthetic generation with subject matter experts. Its annualized revenue run rate now stands at $375 million.

The growth mirrors a broader boom in the AI training data industry. Companies like Mercor, Handshake, and Micro1 have similarly surged, with Mercor hitting $2 billion in gross annualized revenue earlier this year. Unlike those firms that primarily sell human labor, Snorkel's approach keeps expert payments in cost of goods sold rather than headline-generating revenue.

OpenAI extends its Daybreak cyber defense program to Ukraine

The AI lab announced it will offer the Ukrainian government access to advanced AI models for identifying software vulnerabilities and testing fixes on civilian infrastructure.

The AI lab announced it will offer the Ukrainian government access to advanced AI models for identifying software vulnerabilities and testing fixes on civilian infrastructure.

The program was announced by Dmytro Kushneruk, Consul General of Ukraine in San Francisco, and Sasha Baker, OpenAI's Head of National Security Policy, on the sidelines of the UN General Assembly. Ukraine's CERT-UA handled nearly 6,000 cyber incidents in 2025 alone, including attacks on hospital systems, the energy sector, and telecommunications. Daybreak provides cyber defenders with tools to review older software, investigate suspicious activity, validate vulnerabilities, and test patches.

OpenAI has already extended the program to European partners including France, Germany, and Poland. The EU's ENISA agency used the models to find and fix vulnerabilities across EU institutions, while Poland's CERT Polska discovered six vulnerabilities in third-party router software that have since been patched.

OpenAI launches GPT-6 Sol and Luna at half the price of last year’s models

Updated Sol and Luna models arrive alongside a dramatic price cut, intensifying a broader industry price war sparked by Anthropic’s simultaneous Claude Opus 5.5 release.

Updated Sol and Luna models arrive alongside a dramatic price cut, intensifying a broader industry price war sparked by Anthropic’s simultaneous Claude Opus 5.5 release.

OpenAI expanded the GPT-6 generation with new versions of its Sol and Luna models, available in ChatGPT Work, Codex, and the API. GPT-6 Sol targets complex tasks like coding, while Luna handles high-volume clerical work such as summarizing documents. Both ship at half the cost of the GPT-5.6 equivalents, a reduction OpenAI attributes to improved caching and inference efficiency. GPT-6 Luna, at $0.10 per million input tokens, is one of the cheapest models OpenAI has ever released.

OpenAI claims GPT-6 Sol makes about half as many mistakes as its predecessor on internal factuality evaluations. The models launched roughly 90 minutes after Anthropic released Claude Opus 5.5 at a 20 percent discount, underscoring the intense competition between the two labs as they fight for enterprise and developer mindshare.

Xapien brings AI due diligence to ServiceNow workflows

The due diligence platform launches as a native application on the ServiceNow AI Platform, embedding automated background checks directly into third-party risk and onboarding workflows.

The due diligence platform launches as a native application on the ServiceNow AI Platform, embedding automated background checks directly into third-party risk and onboarding workflows.

Xapien, which closed a $56 million Series B round less than two weeks prior, now delivers fully sourced, human-grade due diligence on any person or organization directly inside ServiceNow. The app is available in the ServiceNow Store with free trials for customers. It enriches out-of-the-box risk scores in Third-Party Risk Management and is configurable to other modules. The integration targets risk and compliance teams, procurement, and business development teams by running full sourced background checks without requiring users to leave the system they already use.

ClickHouse appoints former ServiceNow CFO Mike Scarpelli to its Board

The veteran finance executive who guided Snowflake, ServiceNow, and Data Domain through their IPOs joins the board as ClickHouse surpasses $350 million in annualized revenue.

The veteran finance executive who guided Snowflake, ServiceNow, and Data Domain through their IPOs joins the board as ClickHouse surpasses $350 million in annualized revenue.

Scarpelli, who recently served as CFO of Snowflake, will serve as an independent director and chair the board's Audit Committee. He joins ClickHouse as it grows from roughly $200 million in run-rate revenue at the end of 2025 to over $350 million today. The database company, valued at $15 billion in its Series D earlier this year, now serves more than 4,000 customers including DoorDash, Meta, Tesla, Cisco, and Visa, and employs nearly 800 people across 27 countries with plans to reach 1,000 by year-end.

ServiceNow stock gets analyst upgrade as AI governance narrative strengthens

An enterprise software analyst raised the price target on ServiceNow, citing its deepening position as the governance layer for enterprise AI adoption.

An enterprise software analyst raised the price target on ServiceNow, citing its deepening position as the governance layer for enterprise AI adoption.

The stock inched higher on Monday following the upgrade. The move reflects growing investor confidence in ServiceNow's strategy of embedding itself as the central control tower for how large organizations deploy AI across IT, HR, and customer service workflows. Analyst commentary highlights that once enterprises choose a platform for AI deployments, the decision is hard to reverse, and ServiceNow is currently winning that decision against competitors including Salesforce, which trades at a fraction of ServiceNow's valuation multiple.