Anthropic’s Nasdaq debut draws attention to the AI market but does not settle concerns about its durability. Shares briefly reached $115—more than 38% above the offering price—before closing 17% higher, putting the company’s fully diluted valuation above $2 trillion. They fell further in premarket trading Friday. The debut drew focus from Nvidia’s strong earnings, while another account says the IPO met Wall Street expectations without reviving broader investor interest in AI stocks. Commentators warn that a sell-off could put AI-driven wealth gains at risk, and questions remain about Anthropic’s business model. One article also references market unease around a sex and spy scandal involving Nvidia CEO Jensen Huang; another reports Hunterbrook Media’s claim that AMD’s ties to Chinese entities could face more U.S. regulatory scrutiny. Hunterbrook says it is shorting AMD.
Cybersecurity reports describe threats ranging from automated attacks and stolen credentials to software flaws, supply-chain abuse and outages. One roundup cites a Maileva breach affecting 22 million people and Uber driver records exposed after Fireworks AI credentials were compromised. It recommends phishing-resistant multifactor authentication, least-privilege access and promptly removing former employees’ credentials. Separately, a European Union official said an AI-powered attack targeted a key EU institution in October and was addressed by cybersecurity experts for almost a month. Some digital tools remained offline, but officials have not disclosed the perpetrator, affected systems or extent of damage. The bloc described it as its first reported cyberattack using an AI tool.
Security vendor Cato Networks has introduced AgentSense, developed by Cato AI Labs to detect prompt injection and stop unauthorized actions by autonomous AI agents. Its Untrusted Influence Detection separates verified instructions from potentially harmful content, while Drift Detection checks whether agents remain aligned with their original goals. Cato is offering $10,000 to each of the first three participants who bypass the system in a security challenge.
AI providers are also experimenting with charging for managed agents. WhiteMatter, formerly QuilrAI, says it manages 2.25 million agents, projects 5.5 million by year-end and has grown revenue fivefold in six months. Nestack charges $199 to $499 per agent monthly, billing only for agents in production. Dataiku combines an annual platform fee with per-agent monitoring charges but has not disclosed its rate. Microsoft, by contrast, uses per-user licenses and usage-based charges: it reports more than 50 million agents registered with Agent 365 but does not bill per agent.
Coinbase CEO Brian Armstrong says the company is building financial accounts and payment rails for AI agents, and predicts agents will eventually make 99.9% of transactions. He also argues that teams of open-weight models could offer an alternative to a future dominated by a few frontier AI labs. In wealth management, Humanity Labs CEO Andrei Pop says AI could help advisory firms deliver more to every client, while requiring changes to organizations and employee responsibilities.
In Japan, Fujitsu and FamilyMart plan a limited trial using Fujitsu’s Kozuchi causal AI and large language model technology to investigate why shoppers choose lower-carbon products. The trial runs from November 2026 to February 2027, with marketing tests planned for January at selected Famima!! stores. It starts with Convenience Wear clothing, including innerwear made from recycled polyester, and tests the effects of environmental information and shopper benefits. The companies report no sales increase; possible later applications to food and daily necessities have no confirmed launch or timetable.
Key Takeaways
- Anthropic’s Nasdaq shares briefly rose above $115, closed 17% above the offering price and implied a fully diluted valuation above $2 trillion; the stock fell further in premarket trading Friday.
- Anthropic’s debut did not revive broader AI-stock interest, while market commentary flags bubble risks and questions about its business model. Nvidia’s earnings were strong.
- Hunterbrook Media says AMD’s China-related ties could face greater U.S. regulatory scrutiny and says it is shorting the company.
- A reported AI-powered attack targeted an EU institution in October; some digital tools remained offline after nearly a month of response, with the attacker and damage undisclosed.
- A cybersecurity roundup cites a Maileva breach affecting 22 million people and Uber driver data exposed through compromised Fireworks AI credentials.
- Cato Networks launched AgentSense to detect prompt injection and agent drift, offering $10,000 to each of the first three people to bypass it.
- WhiteMatter reports 2.25 million managed agents and fivefold revenue growth in six months; Nestack charges $199–$499 monthly per production agent.
- Microsoft reports more than 50 million agents registered with Agent 365 but charges through per-user licenses and usage, not per-agent billing.
- Coinbase CEO Brian Armstrong predicts agents will eventually make 99.9% of transactions and says the company is building accounts and payment rails for them.
- Fujitsu and FamilyMart will test lower-carbon product marketing at selected Japanese stores from January, within a November 2026–February 2027 trial; no sales increase has been reported.
Cybersecurity threats this week include AI attacks and credential theft
This weekly roundup describes threats including automated attacks, stolen credentials, software flaws, supply chain abuse and service outages. It says attackers can use exposed services, excessive permissions and stolen logins to gain access quickly. The article also lists data exposures, including a breach affecting 22 million people at Maileva, and an attack that exposed Uber driver records through compromised Fireworks AI credentials. It recommends steps such as phishing resistant multifactor authentication, least privilege access and prompt removal of former employees’ credentials.
Cato Networks launches AgentSense to guard autonomous AI agents
Cato Networks launched AgentSense, a security tool developed by Cato AI Labs to detect prompt injection and prevent unauthorized actions by autonomous AI agents. Its Untrusted Influence Detection feature separates verified instructions from potentially harmful content, while Drift Detection checks that agents stay aligned with their original goals. Cato is offering $10,000 to each of the first three participants who bypass the system in a security challenge. The company says the tool could help organizations and partners provide AI monitoring and governance services.
AI powered cyberattack targeted a key European Union institution
An official confirmed on Thursday, November 19, that cybersecurity experts had been dealing with an AI generated attack on a key European Union institution for almost a month. Some of the bloc’s digital tools remained offline. The official did not identify who was behind the attack or provide further details. It was the first time the 27 member bloc reported a cyberattack using an AI tool.
AI powered cyberattack targeted the European Union in October
An official in Brussels said a cyberattack using AI powered software targeted the European Union in October. The attack affected some of the bloc’s digital systems, but the official did not say which systems were involved or what damage occurred. It was detected in October, and the official provided no further details. The official was not authorized to be named.
Anthropic IPO draws attention but does not revive AI stocks
Anthropic’s leaders celebrated the company’s market debut at Nasdaq on Thursday. The article says the IPO met Wall Street’s expectations but did not revive investor interest in AI stocks. It also refers to market unease following a sex and spy scandal involving Jensen Huang. The provided article gives no further details about the IPO or the market reaction.
Fujitsu and FamilyMart test AI to understand green purchases
Fujitsu and FamilyMart will use Fujitsu’s Kozuchi causal AI and large language model technology to study why shoppers choose products with a lower carbon impact. The trial runs from November 2026 to February 2027, with marketing tests planned for January at selected Famima!! stores in Japan. It will begin with Convenience Wear clothing, including innerwear made with recycled polyester, and test how environmental information and shopper benefits affect buying decisions. The project is a limited trial, not a nationwide rollout, and the companies have not reported a sales increase. They may later apply findings to food and daily necessities, but no launch or timetable has been confirmed.
AI market sell off puts recent wealth gains at risk
A recent stock market sell off threatens some of the wealth created during the AI rally. The rise in companies such as Anthropic and Google’s DeepMind helped drive market growth in recent years. A hedge fund manager warned that AI wealth could become a bubble that hurts the wider market if it bursts. A market analyst said AI supports innovation and growth, but also brings risks, and the sell off highlights how quickly promising trends can change.
AI could help advisory firms deliver more to every client
AI could change how wealth management firms compete by allowing them to provide more value to every client. In a discussion sponsored by Humanity Labs on November 20, 2026, founder and CEO Andrei Pop described how AI may expand what advisory firms can offer beyond saving time. He also discussed organizational changes firms may need to make. These changes include rethinking employee responsibilities as AI takes on more work.
AI firms move toward charging for each managed agent
Some AI vendors are introducing pricing based on each agent they manage or monitor, rather than relying only on employee licenses or usage fees. WhiteMatter, a San Francisco startup formerly called QuilrAI, manages 2.25 million agents and projects 5.5 million by year end; it says its revenue grew fivefold in the past six months. Nestack charges $199 to $499 per agent per month, depending on the agent’s role and support needs, and bills only for agents in production. Dataiku combines an annual platform fee with a monitoring charge for each agent, though it has not published the per-agent rate. Microsoft uses per-user licenses and usage-based charges, and has more than 50 million agents registered with Agent 365, but does not bill per agent.
Coinbase chief sees AI agents handling nearly all transactions
Coinbase CEO Brian Armstrong spoke with Aoyon Ashraf at the TomorrowX Summit about the company’s plans for AI agents. He explained why Coinbase is building financial accounts and payment rails for these agents. Armstrong believes that 99.9% of all transactions will eventually be made by agents rather than people. He also discussed why open weight models working in teams could counter a future dominated by a few powerful frontier AI labs.
Anthropic’s $2 Trillion IPO Raises Questions About AI Boom
Anthropic began trading on the Nasdaq on Thursday, and its shares briefly rose to $115, more than 38% above the offering price. The stock ended the day up 17, putting its fully diluted valuation above $2 trillion, but fell further in premarket trading Friday. The IPO drew attention away from Nvidia, whose earnings were strong, but the article says Anthropic’s debut may signal that the four year old bull market is becoming stretched. It also notes concerns about Anthropic’s business model and a Hunterbrook Media report that AMD’s ties to Chinese entities could face more US regulatory scrutiny. Hunterbrook is shorting AMD.
Sources
- AI Attacks, Credential Theft, and Critical Flaws Define the Week
- Cato Networks introduces AgentSense to secure autonomous AI agents
- A cyberattack using artificial intelligence targeted the EU in October, official says
- A cyberattack using artificial intelligence targeted the EU in October, official says
- Anthropic’s Mega IPO Lives Up to Wall Street Hype but Fails to Revive AI Trade
- Fujitsu FamilyMart AI Trial Tests What Drives Purchases
- The ripple effects of AI wealth
- Beyond Efficiency: AI’s Bigger Opportunity for Advisory Firms
- AI Vendors Ditch Seat Licenses for Per-Agent Pricing
- Brian Armstrong on bitcoin’s bottom, AI trading agents and Coinbase’s next act
- Morning Bid: AI’s last stand?
Comments
Please log in to post a comment.