Meta's Muse AI Agent Causes Concern, Analysts See Potential

Key Takeaways

• <ul><li>Meta's Muse AI agent caused a user to want to delete the app after a strange message</li><li>Analysts see Muse as a potential factor in how investors judge Meta's AI spending</li><li>Anthropic is under pressure to delay its IPO over AI safety and governance concerns</li><li>NYC Comptroller's pension funds hold about $7 million exposed to Anthropic</li><li>Prosecutors from eight U.S. states raised concerns about Anthropic's IPO timing</li><li>Supermicro expands Open Compute Project portfolio to speed up AI data center deployment</li><li>SentinelOne starts Strong Buy coverage with a $34 price target and 36% upside</li><li>AI-powered ETFs manage about $30 million, with some beating the S&P 500</li><li>Google has trained over 80,000 educators in India using Gemini tools</li><li>Gartner warns over $3.5T in AI infrastructure investments could be written off by 2030</li></ul>

I stopped using Meta Muse after a strange AI message

Ugo Nwune tried Meta's Muse AI agent for 3 days. The agent sent him a strange message that made him want to delete the app. He decided to stop using Muse completely. The incident highlights concerns about AI agent behavior. The experience left him uneasy about continuing to use the tool.

Wall Street may overlook Meta's growing AI business

Wall Street may be underestimating Meta's AI business potential. BofA analyst Justin Post says Meta launched new frontier AI models from Superintelligence Labs. The models are praised for speed compared to OpenAI's offerings. Meta's Muse AI agent was released in September. Deutsche Bank analyst Benjamin Black sees potential for Muse to change investor views on AI spending.

Meta Muse could shift views on AI spending this earnings

Meta's Muse AI agent may change how investors think about AI capital expenditure during the earnings season. The agent was released in September and is gaining attention. Business Insider reports that Muse could influence investor expectations about Meta's AI infrastructure spending. The story is available exclusively to Business Insider subscribers.

Anthropic asked to delay IPO over AI security fears

New York City Comptroller Mark Levine asked Anthropic to postpone its IPO due to AI security concerns. Levine cited safety and governance worries in a letter to the company. He oversees pension funds for over 750,000 people with $330 billion in assets. About $7 million of those assets are exposed to Anthropic. Prosecutors from eight U.S. states also raised concerns. Anthropic's IPO could increase the city's automatic exposure through stock index inclusion.

NYC Comptroller urges Anthropic to delay IPO for AI safety

New York City Comptroller Mark Levine called on Anthropic to delay its IPO and focus on AI safety. Levine said the company should address AI threat concerns and build proper governance before going public. He referenced a September 12 report and asked Anthropic to detail its audit plans and timeline. Levine manages pension funds with over $325 billion in assets. Anthropic did not immediately respond to requests for comment.

Supermicro expands open hardware for faster AI data centers

Supermicro is growing its Open Compute Project portfolio with integrated AI racks, networking, power, and cooling. The company is packaging OCP racks, 48-volt power, and liquid cooling into validated systems. This aims to reduce complexity and speed up AI cluster deployment. It targets rising demand for AI and data analytics. Supermicro also offers design, implementation, and support services. The solutions are built to scale and flex with customer needs.

SentinelOne starts Strong Buy coverage with $34 target

SentinelOne began coverage with a Strong Buy rating and a $34 price target, showing 36% upside. The company is growing beyond endpoint protection into AI security, cloud, and data analytics. Non-GAAP operating margins are expected to reach 18% by 2028. Adjusted EPS is estimated at $0.59. Key risks include margin execution, competition, and stock-based compensation.

AI stock funds use bots to pick investments

Some fund managers are launching AI-powered ETFs that use bots to pick stocks. Fewer than half a dozen ETFs rely only on AI, managing about $30 million of the industry's nearly $17 trillion. Early results show some AI funds beating the S&P 500. One fund launched in August is up 2.7%, compared to 1.1% for the S&P 500. Experts note short records and higher fees may slow investor interest.

Google expands AI teacher training in India

Google is growing its Google AI Educator Series with partnerships in Bihar, Madhya Pradesh, and Sikkim. Over 80,000 educators have been trained since May. The program teaches teachers to use tools like Gemini and Gemini Notebook for lesson plans and schoolwork. Training is in English now, with more languages coming. Google also supports 100 Atal Tinkering Labs with a Gemini-powered app called ATL Saathi.

Gartner warns AI infrastructure risks hit neocloud stocks

Gartner warned that over $3.5T in AI infrastructure investments could be written off by 2030. This has raised worries about AI boom profits and pushed neocloud stocks lower. High capital spending, rising financing costs, and weak revenue forecasts threaten provider profits. Over 80% of neocloud providers are expected to stop operating independently due to costs and limited differences.

Typeface Research: Brands Sacrifice Authenticity for AI Visibility

A survey of over 500 marketing leaders found that brands are sacrificing quality and authenticity to get cited by AI. The Typeface Signal Report shows 62% say content quality is being traded for AI citations, and 61% say the same about authenticity. Most leaders expect AI to become a top discovery channel by 2030, but many teams still work in silos. The report urges brands to build a shared foundation so AI visibility does not erase what makes them unique.

Open-Weight Models Match Closed Models on Security Benchmark

A study of CTI-REALM tested whether open-weight language models can handle security agent tasks as well as costly closed models. Researchers evaluated 26 models, including 19 open-weight and 7 closed ones, on a threat-intelligence task. The results showed open-weight models achieved similar performance while being cheaper and more transparent. This finding supports using open-weight models for security applications and calls for more research.

Sources

• My Meta Muse agent sent me a strange message. I stopped using it.
• What Wall Street may be forgetting about Meta's AI business
• Meta's Muse could change investors' views on AI capex this earnings season
• Anthropic asked to delay IPO over AI security concerns
• NYC Comptroller Urges Anthropic to Delay IPO and Focus on AI Safety
• Supermicro bets on open hardware to speed AI data center construction
• SentinelOne: Why I Believe The AI Security Rally Has More Room To Run
• AI-powered stock funds bet bots can pick stocks faster and better than humans
• Google Expands AI Teacher Training In India Through Partnerships With Three States
• Most neocloud stocks rated Hold or worse as Gartner flags AI infrastructure risks
• The AI Visibility Gap: Brands Are Trading Away the Authenticity That Made Them Worth Finding, New Typeface Research Finds
• Unconstrained CTI-REALM with Open-Weights