Anthropic reportedly raises $81 billion at $1.5 trillion valuation

US-China technology tensions are weighing on Chinese and Hong Kong AI shares. Several reports describe the CSI300 and Shanghai Composite down about 0.4% near 16-month lows, while Hong Kong’s Hang Seng Tech index fell between 0.6% and 1.5%, with one report putting it at a more-than-two-year low. The Trump administration plans to block imports of Chinese data-centre equipment, and US officials have considered restrictions on Chinese optical transceivers. Nvidia’s outlook assumes no data-centre chip sales to China. US-dependent optical-component makers Zhongji Innolight and Eoptolink have fallen more than 70% from their peaks. China’s STAR Chip Index rose in some reports, as investors saw a possible benefit for local chipmakers; gold shares fell nearly 5% in reports citing higher US rate expectations. The selling came ahead of a meeting between Xi Jinping and Donald Trump, with comments by former Nvidia CEO Jensen Huang also cited as a concern.

Microsoft and Salesforce announced Copilotforce at Microsoft Ignite on November 17, 2026, bringing Salesforce customer data and sales actions into Microsoft Copilot and Teams. Its plugin, with more than 38 sales skills, is in public preview; sellers can find customer information, prepare for meetings, review deals and draft follow-ups while using existing Salesforce permissions. Sales Cloud in Teams remains in a pilot with enterprise customers, with an open beta planned for later this year. Salesforce Data 360 and Microsoft Fabric also support two-way, zero-copy data sharing, allowing queries without duplicate datasets. Accenture is the design and implementation partner and a pilot user. The integration may give small businesses faster access to customer insights, though staff training takes time and resources.

Banks continue to block outside AI agents from accessing customer accounts over security concerns. JPMorganChase and Bank of America have taken a restrictive approach, although Bank of America is also exploring ways to permit agents. Bankers at The Clearing House’s Annual Conference said policies could change as customers adopt personal agents and banks establish controls. Separately, guidance on preparing for Office of the Comptroller of the Currency AI reviews says the framework creates no new rules or supervisory expectations. Polsinelli advises banks to inventory AI tools and document their capabilities, data access and human-approval requirements; assess customer risks, privacy and security; review vendors and contracts; and prepare to stop agents, revoke access, investigate incidents and limit harm while maintaining service.

Joanne Wright and Yogi Goyal argue that enterprise AI needs agreement across the C-suite and a reusable operating model covering intelligence, action, operations and control. A COO may guide the work but must build consensus across business units. They say organizations often stall after setting a vision, when functions begin separate AI projects. Technology can turn procedures into agent use cases, business cases and plans in hours, but deciding what to build remains challenging.

AI policy and safety are also moving into public debate. Democrats preparing to take control of Congress in January may clash with President Donald Trump over AI, data centres and tech companies’ responsibility for harm. An NBC News exit poll found 56% of voters considered AI and data centres a major or crucial factor in their vote, while 76% opposed a new data centre in their community. Proposals include requiring data centres to pay for added electricity infrastructure, addressing AI-related harm, pausing advanced AI and taxing some AI windfall profits. Democrats could pursue narrower bills, attach rules to spending measures or hold hearings, but Trump could veto legislation.

In consumer fraud, Surrey Police and Crime Commissioner Lisa Townsend warned that AI voice-cloning scams have become common in recent months. She demonstrated an AI copy of her voice and said scammers may need only three seconds of a person’s voice to generate full sentences. UK Finance figures published in June recorded more than four million fraud cases involving stolen money—up 11% from 2024 and 31% from 2023. Residents said scams are becoming more sophisticated and can be easy to fall for when people feel overwhelmed.

Researchers’ Evidence-based Credit Assignment method aims to help search-augmented language models learn which retrieved documents support an answer. Across seven Search QA benchmarks, ECA-Horizon improved strict macro exact match over outcome-only GRPO by 1.03 percentage points on Qwen2.5-3B-Instruct and 4.80 points on Qwen2.5-7B-Instruct; zero-shot tests also showed transfer to multi-hop questions. Meanwhile, Anthropic, maker of the Claude chatbot, reportedly raised $81 billion ahead of a stock market debut, reaching a reported $1.5 trillion valuation. The report notes questions about whether AI’s rapid growth signals a bubble. A separate market segment featured Prime Capital Financial portfolio manager Clayton Allison discussing the AI trade amid controversy involving Nvidia, but provided no further detail on his views.

Key Takeaways

  • US curbs on Chinese data-centre equipment and potential optical-transceiver restrictions are pressuring China and Hong Kong AI shares; Nvidia’s outlook assumes no data-centre chip sales to China.
  • Zhongji Innolight and Eoptolink are more than 70% below their peaks, while China’s STAR Chip Index rose in some reports as investors considered potential gains for local chipmakers.
  • Microsoft and Salesforce launched Copilotforce at Ignite on November 17, 2026; its Copilot plugin offers more than 38 Salesforce sales skills and is in public preview.
  • Salesforce Data 360 and Microsoft Fabric provide two-way, zero-copy data sharing; Accenture is the design and implementation partner and a pilot user.
  • JPMorganChase and Bank of America block outside AI agents from banking systems, though Bank of America is exploring ways to allow them.
  • OCC AI-review guidance adds no new rules; recommended bank preparations include tool inventories, risk and vendor reviews, human controls, and tested incident-response plans.
  • Democratic proposals include making data centres pay for added electricity infrastructure and measures addressing AI harms; an NBC News exit poll found 76% opposed a new data centre in their community.
  • Surrey’s police and crime commissioner warned that scammers may clone a voice from just three seconds of audio; UK Finance reported more than four million fraud cases involving stolen money.
  • ECA-Horizon improved strict macro exact match by 1.03 points on Qwen2.5-3B-Instruct and 4.80 points on Qwen2.5-7B-Instruct across seven Search QA benchmarks.
  • Anthropic, maker of Claude, reportedly raised $81 billion at a $1.5 trillion valuation ahead of a stock market debut; a separate report discussed Nvidia-related AI-trade controversy without detailing the investor’s views.

China and Hong Kong AI shares fall as US tech curbs loom

Chinese and Hong Kong stocks fell on Wednesday as investors sold AI-related shares amid signs the United States may tighten technology limits on China. By midday, the CSI300 and Shanghai Composite were each down about 0.4%, while Hong Kong’s Hang Seng Tech index fell 1.5%. The Trump administration plans to restrict imports of Chinese data-centre equipment, and Nvidia’s outlook assumes no sales of its data-centre chips to China. Chinese optical-component makers Zhongji Innolight and Eoptolink fell further and were more than 70% below their peaks, while the STAR Chip Index rose 0.5%.

China AI shares fall as investors worry about tighter US restrictions

Chinese and Hong Kong stocks fell as investors sold Chinese AI hardware exporters and AI hyperscalers ahead of a meeting between Xi Jinping and Donald Trump. China’s stock benchmark was near a 16-month low, while comments from former Nvidia CEO Jensen Huang added to investor concerns about US-China technology ties. The article says the United States has tightened technology rules and export controls on Chinese companies, raising concerns about their access to the US market. It also warns that these restrictions could weigh on Chinese technology stocks.

China tech shares slide as US restrictions threaten AI companies

Chinese and Hong Kong shares fell as investors sold AI hardware exporters and hyperscalers amid signs of tougher US technology curbs ahead of a meeting between Xi Jinping and Donald Trump. The CSI300 and Shanghai Composite each fell about 0.4% by midday, near their lowest levels since mid-2025. The Trump administration plans to block imports of Chinese data-centre equipment, and Nvidia’s outlook assumes no data-centre chip sales to China. Zhongji Innolight and Eoptolink, which rely heavily on the US market, were down more than 70% from their peaks. Hong Kong’s Hang Seng Tech index fell 1.5%, while China’s STAR Chip Index gained 0.5% as investors bet local chipmakers could benefit from US restrictions; gold stocks fell nearly 5% amid higher expectations of US rate hikes.

China and Hong Kong technology shares hit by US curb concerns

Chinese shares traded near 16-month lows as investors sold AI-related companies amid signs of tougher US technology restrictions. By midday, the CSI300 and Shanghai Composite were each down about 0.4%, while Hong Kong’s Hang Seng Tech index fell 1.5%. The Trump administration plans to block imports of Chinese data-centre equipment, and US officials have considered banning new models of Chinese optical transceivers. Zhongji Innolight and Eoptolink, which depend heavily on US business, were more than 70% below their peaks. China’s STAR Chip Index rose 0.5% as investors expected local chipmakers to benefit from tighter US limits, while gold stocks fell nearly 5%.

US chip curbs weigh on China AI and technology shares

Chinese shares struggled near 16-month lows and Hong Kong stocks fell as investors sold AI hardware exporters and hyperscalers ahead of a meeting between Xi Jinping and Donald Trump. The Trump administration plans to block imports of Chinese data-centre equipment, and Nvidia’s outlook assumes no data-centre chip sales to China. Chinese optical-parts makers Eoptolink and Zhongji Innolight, which earn most of their revenue from the United States, were down more than 70% from their peaks. Hong Kong’s Hang Seng Tech index fell 0.6% to its lowest level in more than two years, with Alibaba and Tencent also reaching multi-month lows. China’s STAR Chip Index rose 2.5%, while gold stocks fell 4% amid higher expectations of US rate hikes.

Microsoft and Salesforce bring sales tools into Copilot

Microsoft and Salesforce announced Copilotforce at Microsoft Ignite on November 17, 2026. It brings Salesforce customer data and sales actions into Microsoft Copilot, with a plugin featuring more than 38 sales skills now in public preview. Sellers can review deals, find customer information, and draft follow-ups while working in Microsoft tools, using their existing Salesforce permissions. Sales Cloud in Teams remains in a pilot with enterprise customers, while an open beta is planned for later this year. The companies also offer data sharing between Salesforce Data 360 and Microsoft Fabric, and Accenture is the design and implementation partner and a pilot customer.

Salesforce and Microsoft connect sales data with AI tools

Salesforce and Microsoft announced Copilotforce at Microsoft Ignite, bringing Salesforce customer information and sales workflows into Microsoft Copilot and Teams. Sellers can use the integration to prepare for meetings, review sales opportunities, and take CRM actions based on information from Teams and Outlook. A two-way zero-copy integration between Salesforce Data 360 and Microsoft Fabric lets organizations query data without keeping duplicate datasets. Accenture is the design partner and an early user. The article says small businesses may benefit from faster access to customer insights, while noting they may need to invest time and resources in staff training.

Banks weigh security as customers bring AI agents to banking

Banks are largely blocking outside AI agents from accessing their systems, raising questions about the security controls these tools may need. The article says JPMorganChase and Bank of America have taken a hard line against customer AI agents, though Bank of America is also exploring ways to allow them. Bankers speaking at The Clearing House’s Annual Conference said these policies may change as more consumers become comfortable with personal agents and banks settle on controls.

Banks may ease restrictions on customer AI agents

Many banks are blocking customers from using outside AI agents in their banking systems because of security concerns. JPMorganChase is blocking these agents, and Bank of America has taken the same stance while also exploring ways to allow them. Bankers speaking at The Clearing House’s Annual Conference said the restrictions may not last as more consumers become comfortable with personal AI agents and banks develop suitable controls.

Building an AI operating model starts with executive agreement

Joanne Wright and Yogi Goyal say companies need agreement across the C-suite to make AI useful across the enterprise. They describe an AI Operating Model as a reusable foundation built around intelligence, action, operations, and control. The COO often helps guide the work, but must build consensus because no single leader can make decisions for every business unit. The authors say companies often stall after agreeing on a vision, when separate functions start building their own AI projects. They add that technology can now turn company procedures into agent use cases, business cases, and plans in hours, while agreeing on what to build remains difficult.

Surrey police chief warns of AI voice cloning scams before Christmas

Surrey Police and Crime Commissioner Lisa Townsend warned that AI voice cloning scams have become common in recent months. She demonstrated the technology using an AI copy of her voice and said scammers may need only three seconds of someone’s voice to create full sentences. Residents Enid Jones and Victoria Oso said scams are becoming more sophisticated and can be easy to fall for when people feel overwhelmed. UK Finance figures published in June showed more than four million cases of fraud involving stolen money, an 11% increase from 2024 and 31% from 2023.

How banks can prepare for OCC reviews of AI

The Office of the Comptroller of the Currency’s AI framework took effect as a guide for reviews, but the OCC says it creates no new rules or supervisory expectations. Polsinelli recommends that banks list their AI tools and record what each tool can do, what data it can access, and which actions need human approval. Banks should assess customer risks, test privacy and security controls, review vendors, and update contracts to cover AI features and data use. They should also prepare plans for AI incidents and test whether they can stop agents, revoke access, investigate activity, limit customer harm, and keep services running.

Democrats weigh AI rules as they prepare to take control of Congress

Democrats are preparing to control Congress in January, setting up a possible conflict with President Donald Trump over AI, data centers, and technology companies’ responsibility for harm. An NBC News exit poll found 56% of voters saw AI and data centers as a major or crucial factor in their vote, and 76% opposed a new data center in their community. Proposed measures include the Bipartisan American Affordability and Jobs Act, which would make data centers pay for added electricity infrastructure, and efforts to address harm caused by AI systems. Arizona lawmakers Yassamin Ansari and Mark Kelly have also proposed a pause on advanced AI and a tax on some AI windfall profits. Democrats may pursue narrower bills, attach rules to spending legislation, or hold hearings, though Trump could veto legislation.

ECA method helps search AI models learn which sources matter

Researchers introduced Evidence-based Credit Assignment, or ECA, to help search-augmented language models learn which retrieved documents support an answer. The method adds a training-only signal that measures how documents change the likelihood of a reference answer, addressing a weakness in outcome-only rewards. The researchers tested two versions, ECA-Step and ECA-Horizon, along with cross-fold group normalization, novelty gating, and search-local token assignment. Across seven Search QA benchmarks, ECA-Horizon raised strict macro exact match over outcome-only GRPO by 1.03 percentage points on Qwen2.5-3B-Instruct and 4.80 points on Qwen2.5-7B-Instruct. Zero-shot tests on MultiHop-RAG also showed strong transfer to multi-hop inference queries.

Anthropic raises $81 billion ahead of stock market debut

Anthropic, the company behind the Claude AI chatbot, raised $81 billion ahead of its stock market debut. The article reports that the company’s valuation reached $1.5 trillion. The flotation comes as questions grow about whether the AI industry’s rapid growth is showing signs of a bubble. Anthropic was founded in 2021, and its seven founders have become multi-billionaires.

Clayton Allison weighs AI trade outlook amid Nvidia controversy

Clayton Allison, a portfolio manager at Prime Capital Financial, joined BNN Bloomberg’s The Close on November 17, 2026. He discussed the market outlook amid controversy involving Nvidia and questions about the AI trade. The report does not provide further details about his views.

Sources

NOTE:

This news brief was generated using AI technology (including, but not limited to, Google Gemini API, Llama, Grok, and Mistral) from aggregated news articles, with minimal to no human editing/review. It is provided for informational purposes only and may contain inaccuracies or biases. This is not financial, investment, or professional advice. If you have any questions or concerns, please verify all information with the linked original articles in the Sources section below.

US-China technology tensions Chinese AI stocks Data center equipment Export restrictions Nvidia Optical transceivers Microsoft Copilot Salesforce Copilotforce Enterprise AI integration Zero-copy data sharing AI agents in banking Banking security OCC AI guidance AI governance AI regulation Data center policy AI voice-cloning scams Search-augmented language models Evidence-based Credit Assignment Anthropic valuation

Comments

Loading...