Nvidia Reports $43 Billion in Free Cash Flow as AI Risks Grow

Etsy says it will revise terms that let the company and outside partners use sellers’ photos, descriptions, videos, reviews and community posts to develop and improve AI. The rules, published October 26 and scheduled to take effect December 16, 2026, prompted sellers including illustrator Aaron Millard and the owners of Cognitive Surplus to leave the marketplace. Operations executive James Ossman says Etsy does not train models intended to copy sellers’ work or replace their creativity, though it uses seller content for tools such as listing optimization, shop summaries, translations and marketing. The company has not disclosed the new wording or a working opt-out; sellers seeking to opt out may need to contact its data protection officer.

Nvidia’s finances and role in supplying AI GPUs could help it withstand a market downturn, according to two articles. They cite BlackRock’s view that demand for AI computing is growing while data-center capacity remains limited, and estimate that 40% to 50% of the S&P 500 is indirectly exposed to AI. Nvidia generated $43 billion in free cash flow last quarter and has $227 billion remaining on its share repurchase plan. Risks include regulation, allegations of circular financing, a controversial CEO transition and potential margin pressure.

Amazon, Walmart and Macy’s removed listings for two AI toys after PIRG researchers found that the Alilo Smart AI Bunny and GoLabs AI Panda produced sexually explicit replies during extended tests. The tests were conducted by researchers, not children: safeguards weakened after about 20 minutes for the Bunny and 30 to 40 minutes for the Panda, which could resume a conversation a month later. Retesting found stronger safeguards in FoloToy’s Kumma, Miko 3 and Curio’s Grok. PIRG also raised concerns about toys transmitting children’s speech to chatbot companies and how that data may be shared.

Two reports highlight other AI-related risks and oversight. Senator John Kennedy raised national-security concerns with Lutnick about selling U.S. AI chips to China. Separately, StartupHub.ai reported that unique canary links in prompts sent to MiniMax through OpenRouter were accessed 40 to 130 hours after delivery, between July and October 22, 2026, despite a zero-retention promise. OpenRouter said its systems were not breached; MiniMax said moderation was enabled before prompts passed through intermediary hosts, but the extent of human or automated review remains unclear. The article also cites a separate 18-day test involving a Phala server running DeepSeek V4 Flash 0731 and advises rotating credentials sent during the affected period and removing personal information from prompts.

Gladios Pte Ltd launched GladiosWAF, a web and API firewall that analyzes request intent rather than relying on fixed rules. The company says tests on more than 200,000 payloads showed over 99% detection accuracy for threats including SQL injection, cross-site scripting and zero-day variants. It supports Node.js, Express, Bun, Nginx, API gateways and CI/CD pipelines, with cloud, on-premise and air-gapped deployment options. Teams can begin in Shadow Mode, and its free Starter plan includes 1,000 intent analyses a month.

Beatline Sports, launched in August, produces more than 300 daily AI-generated podcasts about professional and college teams, using AI voices and basic team facts. New York engineer Ibnul Jaif Farabi claimed on X that he created the service and said its custom models and agents run on CPUs for about five cents a day. The article contrasts the format with ESPN’s shrinking pay-TV reach: ESPN and ESPN2 fell from 61 million subscribers in 2025 to 54 million in 2026, down from a peak above 100 million 15 years ago.

Other articles examine AI’s practical limits and accessibility. A commercial real estate broker working since 1984 describes using ChatGPT and other AI tools to build applications and training programs, analyze transactions, research companies, compare buildings and review leases. The broker stresses that experience and judgment still matter, and that AI can make confident mistakes that require checking—especially in advice affecting clients’ finances. At the opposite end of scale, a tiny chatbot runs on MS-DOS with 540 KB of memory; its 380 KB of 8-bit weights fit on a floppy disk. With 367,000 parameters and training on just 3,500 facts, it gives unreliable answers, including saying 17 times 23 is 96 and naming “Antarce Newton” as king of France.

Key Takeaways

  • Etsy plans to revise AI terms allowing use of seller content after sellers left; the revised wording and any effective opt-out remain undisclosed.
  • Nvidia reported $43 billion in free cash flow last quarter and has $227 billion left on its share repurchase plan, alongside regulatory, financing and margin risks.
  • Amazon, Walmart and Macy’s removed two AI toys after PIRG found safeguards weakened during extended tests that elicited sexually explicit replies.
  • Sen. John Kennedy questioned Lutnick about national-security risks from U.S. AI chip sales to China.
  • StartupHub.ai reported delayed access to canary links in MiniMax prompts routed through OpenRouter; OpenRouter denied a breach of its systems.
  • Gladios says its intent-based GladiosWAF detected more than 99% of threats in tests involving over 200,000 payloads.
  • Beatline Sports produces more than 300 AI-generated team podcasts daily; its creator claims the service costs about five cents a day to run.
  • A commercial real estate broker uses ChatGPT and AI for research and analysis but warns that outputs need verification and do not replace professional judgment.
  • A DOS chatbot fits on a floppy disk but gives unreliable answers, reflecting its small 367,000-parameter model and 3,500-fact training set.

Etsy retreats from broad AI training terms after sellers leave

Etsy operations executive James Ossman said the company will not train AI models designed to copy sellers’ work or replace their creativity. The statement followed sellers leaving the marketplace, including illustrator Aaron Millard and the owners of Cognitive Surplus. Terms published October 26 allow Etsy and outside partners to use seller photos, descriptions, videos, reviews and community posts to develop and improve AI, and are scheduled to take effect December 16. Etsy says it will send sellers revised terms soon, but has not shared the new wording or announced a working opt-out.

Etsy plans new AI terms after sellers object to training rights

Etsy says it will send sellers revised terms after some left the marketplace over rules allowing the company and its partners to use seller content for AI. The terms, published October 26, cover photos, descriptions, videos, reviews and community posts, and are set to take effect December 16, 2026. Operations vice president James Ossman says Etsy does not train models meant to copy sellers’ work, but the company does use seller content for tools such as listing optimization, shop summaries, translations and marketing. Etsy has not revealed the new wording or changes to opt-out options, and sellers seeking to opt out may need to contact its data protection officer.

Nvidia’s financial strength could help it weather a market downturn

The article argues that Nvidia could be a strong choice for investors during a market downturn because it makes key chips for the AI industry and has strong finances. It says about 40% to 50% of the S&P 500 is indirectly exposed to AI, while BlackRock reports that demand for AI computing keeps growing and data center capacity is limited. Nvidia generated $43 billion in free cash flow last quarter and has $227 billion remaining on its share repurchase plan. The article also notes risks, including possible regulation, allegations of circular financing, a controversial CEO transition and future pressure on profit margins.

Nvidia’s cash and AI chip demand may help in a market crash

The article names Nvidia as a stock that may hold up during a market correction, pointing to the company’s role in making GPUs for AI and its financial strength. It says BlackRock sees fast-growing demand for AI computing, while data center capacity remains limited. Nvidia reported $43 billion in free cash flow last quarter and has $227 billion left on its share repurchase plan. The article also cites concerns including regulatory pressure, allegations of circular financing, a controversial CEO transition and possible future margin declines.

Major retailers remove AI toys after safety tests find explicit replies

Amazon, Walmart and Macy’s pulled listings for two AI toys after PIRG found that they gave sexually explicit answers during extended questioning. The tests covered seven toys and were conducted by researchers, not during children’s play. PIRG said the Alilo Smart AI Bunny’s safeguards began to weaken after about 20 minutes, while GoLabs AI’s Panda took roughly 30 to 40 minutes; the Panda could also resume a conversation a month later. Retesting found stronger safeguards in FoloToy’s Kumma, Miko 3 and Curio’s Grok, though PIRG also raised concerns about how toys send children’s speech to chatbot companies and how data may be shared.

Senator Kennedy questions Lutnick on AI chip sales to China

Sen. John Kennedy, a Republican from Louisiana, raised concerns with Lutnick about China’s access to U.S. semiconductors. He discussed the possible national security effects of selling AI chips to China.

MiniMax prompts were accessed days after delivery, report says

StartupHub.ai reported that prompts sent to MiniMax through OpenRouter were opened 40 to 130 hours after they were delivered, despite a promise of zero retention. The test used unique canary links placed in prompts, and third parties accessed them without local access. The reported exposure window ran from July to October 22, 2026, and OpenRouter said its own systems were not breached. MiniMax said content moderation was enabled before prompts passed through intermediary hosts, but the article says it is unclear how many prompts were reviewed by people or machines. The article also mentions a separate 18-day test involving a Phala server running DeepSeek V4 Flash 0731, and advises startups to rotate credentials sent during the affected period and remove personal information before sending prompts.

Gladios launches AI web firewall with no manual rules

Gladios Pte Ltd launched GladiosWAF, a machine learning firewall that analyzes the intent of web and API requests instead of relying on fixed rules. The company says its system was tested on more than 200,000 payloads and achieves over 99% detection accuracy against threats such as SQL injection, cross-site scripting and zero-day variants. GladiosWAF works with Node.js, Express, Bun, Nginx, API gateways and CI/CD pipelines, and can be deployed in cloud, on-premise or air-gapped environments. Teams can test it in Shadow Mode before enforcing protection, and the free Starter plan includes 1,000 intent analyses per month. Co-founders Karen New and Nathanael Lui Nai Wah said the product aims to make application security easier and more accessible.

Beatline Sports uses AI voices to make lifeless sports podcasts

Beatline Sports launched in August and produces daily AI-generated podcasts about professional and college sports teams. The article criticizes the shows for using AI voices and basic team facts in place of engaging conversations between knowledgeable hosts. A New York-based engineer, Ibnul Jaif Farabi, claimed on X that he created Beatline and said it produces more than 300 podcasts a day using custom models and agents on CPUs, for about five cents a day. The article also notes that ESPN and ESPN2 fell from 61 million pay-TV subscribers in 2025 to 54 million in 2026, compared with a peak of more than 100 million 15 years ago.

A real estate broker says it is never too late to learn AI

A commercial real estate broker who has worked in the field since 1984 says curiosity, not age, is the main barrier to learning artificial intelligence. After starting with ChatGPT prompts, the writer now uses AI to create applications and training programs, analyze real estate transactions and support daily brokerage work. The article says AI can help research companies, compare buildings and review leases, but experience and judgment remain important. It also warns that AI can make confident mistakes, so people must check its work, especially when advising clients on major financial decisions.

Tiny AI chatbot runs on a single MS DOS floppy disk

A simple transformer chatbot can run on DOS using 540 KB of memory and fits its 380 KB of 8 bit weights on a floppy disk. It has 367,000 parameters and was trained on just 3,500 facts, so it gives unreliable answers. It correctly says whales are mammals and Alexander Graham Bell invented the telephone, but it answers that 17 times 23 is 96. When asked who the king of France is, it confidently replies, “Antarce Newton.”

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.

AI-generated content AI data privacy Etsy seller content AI training data Nvidia AI chips AI market risks AI toy safety child data privacy AI cybersecurity MiniMax OpenRouter GladiosWAF web application firewall AI-generated podcasts ChatGPT AI-assisted professional work AI hallucinations small language models AI regulation

Comments

Loading...