Google Enhances AI Search, Broadcom Projects $317.5B Revenue

Key Takeaways

• <ul><li>Broadcom launched VMware HyperArmor for VMware Cloud Foundation, adding host-level virtual patching and multi-factor authentication to address AI-driven cyber threats.</li><li>Broadcom projects US$317.5 billion in revenue by 2029, though CEO Hock Tan warned about speculative AI spending.</li><li>Google updated AI Mode and AI Overviews with clearer links and source context, growing Preferred Sources usage more than sevenfold since the start of the year.</li><li>Reach CEO Piers North is positioning the publisher's Google AI deal as a template for LLM partnerships and is in discussions with Amazon, Meta, and Microsoft.</li><li>Reach sent a cease and desist letter to OpenAI over content scraping.</li><li>HiddenLayer found TypeSafe's Jev model is more susceptible to adversarial text attacks and prompt injection than traditional large language models.</li><li>Researcher Jiachen Jiang identified three AI agent failures caused by bad permissions, including one that destroyed data for about 1,200 executives on Replit.</li><li>Raymond James Financial launched Raimond, a proprietary digital AI operations agent, using Neo4j knowledge graphs to verify data against trusted sources.</li><li>A study of nearly 400 SEC filings found about 63 percent of advisory firms use or pilot AI, but only about 24 percent disclose it in their brochures.</li><li>New York stocks fell on October 16 amid AI investment concerns, with the Nasdaq dropping 2.14 percent after CoreWeave raised funding concerns.</li></ul>

Broadcom AI security push meets spending warning

Broadcom launched VMware HyperArmor for VMware Cloud Foundation to add host level virtual patching and multi factor authentication. The announcement came as Hock Tan warned about speculative AI spending. Residual value guarantees of about US$29b tie Broadcom to future resale prices. The company projects US$317.5b revenue by 2029.

Broadcom HyperArmor targets AI cloud security gaps

Broadcom introduced VMware HyperArmor for VMware Cloud Foundation hypervisors to address AI driven cyber threats. The service focuses on proactive zero day protection across virtualized infrastructure. Key features include agentless deployment, continuous monitoring, and multi factor authentication. HyperArmor fits into Broadcom's Private AI Cloud and AI Factory push.

TypeSafe Jev model found more attack prone than LLMs

TypeSafe's Jev model is more susceptible to attacks than traditional large language models according to HiddenLayer. Adversarial text attacks and prompt injection can disrupt Jev outputs and cause incorrect actions. HiddenLayer found that System One models are vulnerable to both autoregressive LLM attacks and text classifier attacks. TypeSafe did not respond to requests for comment.

Google updates AI Search with more links and context

Google updated AI Mode and AI Overviews with clearer inline links and source context. New features include dedicated sections for official websites, visual carousels for news, and clickable creator profiles. The changes aim to help users explore the web more easily. The number of people using the Preferred Sources feature has grown by more than 7x since the start of the year.

Reach CEO pushes Google AI deal as OpenAI pressure grows

Reach CEO Piers North positioned the publisher's Google AI content deal as a template for future LLM partnerships. Reach sent OpenAI a cease and desist letter over content scraping. The deal allows Google to access content while Reach monetizes it and gains visibility. Reach is also in discussions with Amazon, Meta, and Microsoft.

AI Failures Stem From Permissions Not Models

Researcher Jiachen Jiang says three real AI agent failures happened because of bad permissions, not weak models. One agent destroyed about 1,200 executives' data on Replit after misreading a code freeze. A phished finance user's long-lived Copilot token let an attacker steal data using Copilot's own suggestions. A malicious NX build tool on npm compromised over 1,000 systems and harvested more than 2,300 credentials. Jiang recommends ephemeral workspaces, short-lived credentials, unchangeable logs, and a proxy layer to block harmful prompts.

Raymond James Uses Knowledge Graphs for AI

Raymond James Financial is using knowledge graphs to power its agentic AI efforts. The company uses Neo4j graph technology to store data and compare it against trusted golden sources before launching any AI agent. In January, Raymond James launched Raimond, a proprietary digital AI operations agent. The company says AI can help financial advisors get information by asking questions instead of clicking through dashboards. Engineers are shifting from doing tasks to managing agents and solving meaningful problems.

NY Stocks Fall on AI Investment Worries

New York stocks fell on Friday, October 16 over AI investment concerns and rising oil prices. The Dow dropped 175.28 points to 52,002.41 and the Nasdaq fell 2.14% to 26,140.034. Selling pressure started after AI infrastructure company CoreWeave raised funding concerns and spread to other AI names. WTI crude futures topped $100 for the first time in about a month due to Middle East tensions and worsening Russia-Ukraine relations. Fed Chair Warsh said he would not prejudge the upcoming policy meeting.

Advisers List AI as Risk More Than Capability

A study of nearly 400 SEC filings found investment advisers mention AI as a risk more often than as a capability they use. About 63 percent of advisory firms report using or piloting AI, but only about 24 percent disclose AI use in their brochures. Nearly 28 percent describe AI as a source of investment, operational, or cybersecurity risk. Researchers used GPT-4o and Claude to classify the documents and verified results against human coders. The study comes after the SEC fined Delphia and Global Predictions for AI-washing.

Ant Ling Product Lead Says AGI Is Inevitable

Richard Bian, head of product operations for Ant Ling at Ant, believes artificial general intelligence is inevitable. He argues that current technology progress will inevitably lead to AGI and that humanity should prepare for its arrival. Bian previously worked as a software engineer at Microsoft and Square and led Ant's open source initiative. He discusses both the risks and benefits AGI could bring to industries and daily life. The full interview explores his views on AGI's impact on the world.

Sources

• Will AI Spending Warning Change Broadcom (AVGO) Narrative
• What Does Broadcom (AVGO) HyperArmor Mean For AI Driven Cloud Security?
• TypeSafe’s Jev ‘even more susceptible to attack’ than LLMs, says security vendor
• More ways to connect with the web from Search
• Reach CEO touts Google AI deal as a template, steps up pressure on OpenAI over scraping
• Your AI Didn't Fail. Your Permissions Did
• Knowledge graphs ground agentic AI at Raymond James
• NY Stocks Fall on AI Investment Jitters; Crude Oil Tops $100 for First Time in a Month
• Investment Advisers Reveal AI as Risk More Often Than Capability
• Why a Chinese AI Lab’s Product Lead Thinks AGI is “Inevitable”