Cybersecurity stocks saw mixed action recently as AI security concerns created both opportunities and headwinds. Zscaler fell about 8 percent while Palo Alto Networks and Okta each dropped roughly 3 percent, though CrowdStrike declined only about 1 percent. Analysts note that incidents involving AI models breaching company systems are driving demand for better security, even as stretched valuations may cap further gains. CrowdStrike trades at 177 times forward earnings compared to Palo Alto's 94 times, but CrowdStrike's revenue grew 50 percent year over year versus 30 percent for Palo Alto.
CrowdStrike is doubling down on AI agent security with two new products called Falcon Guardian and SafeMind. The company points to a striking finding: a Fortune 500 customer discovered 18,000 AI agents running on its endpoints despite only 300 being approved. SafeMind reportedly improves threat detection by 29 percent, and CrowdStrike believes the AI security market could reach $215 billion by 2034.
Meanwhile, Beazley launched new cyber insurance offerings tailored to AI risks, including an industry-first AI Voluntary Shutdown solution and an AI Regulatory Defence and Penalties cover. These products help organizations pause malfunctioning AI systems and respond to fines or investigations stemming from unintentional AI misuse.
On the legal front, startup DaVoice sued Perplexity AI in California federal court, accusing it of stealing trade secrets related to wake word technology. DaVoice, which has worked with Volkswagen and Samsara, claims Perplexity took confidential source code and training data after a collaboration. A Perplexity spokesperson called the lawsuit baseless and said existing agreements protect its right to build similar products.
Google DeepMind senior engineer Robert O'Callahan resigned over concerns that AI is advancing too quickly. His team worked on chip design tools meant to make AI faster and cheaper, but he concluded that would only accelerate development further. He plans to keep maintaining his open-source debugging tools Pernosco and rr.
In other AI-related developments, Nubank used a simulator called Snowglobe to test customer experience AI agents before deployment. Synthetic customers interacted with agents across over 16,000 conversations, and the simulation-guided changes improved transactional net promoter score by 36.69 points and self-service rate by 8.82 percentage points in live A/B tests.
Federal Reserve Bank of Richmond President Tom Barkin said companies are delaying hiring as they weigh massive AI investment plans, with nearly $700 billion in AI investments announced in a single week earlier this year. Separately, NY Fed Chief Risk Officer Mihaela Nistor called for banks to stress test operational risk from AI, warning that AI thinning long-term leadership pipelines quietly weakens the development of experienced judgment without showing up in loss events.
Research on AI in finance also produced interesting results, where a frozen statistical referee judged AI-proposed investment factors on CSI 500 data and admitted far fewer weak factors than leaky alternatives, though certified true factors waited about 500 trading days on average. In Australia, Bill Ackman argues AI could change how interest rates work, highlighting COSOL with a market cap of A$35.5 million and CAR Group at A$8.2 billion as stocks worth watching.
On the semiconductor side, four chip stocks were recommended for October returns: Analog Devices, SMTC Corporation, Taiwan Semiconductor Manufacturing Company, and ASMIY. The global semiconductor market is projected to grow at a 10.3% compound annual growth rate from 2023 to 2028, driven by AI and machine learning demand.
Key Takeaways
- CrowdStrike launched Falcon Guardian and SafeMind to target AI agent security, with the AI security market projected to reach $215 billion by 2034
- Cybersecurity stocks fell Friday after a strong early week rally, with Zscaler down 8% and Palo Alto and Okta each down about 3%
- DaVoice sued Perplexity AI for allegedly stealing wake word trade secrets, including source code and training data
- Beazley introduced AI Voluntary Shutdown and AI Regulatory Defence insurance covers to address AI operational risks
- Google DeepMind senior engineer Robert O'Callahan resigned over concerns that AI chip design tools would accelerate AI development too fast
- Nubank's Snowglobe simulator tested AI agents across 16,000 synthetic conversations, improving net promoter score by 36.69 points in live tests
- Nearly $700 billion in AI investments were announced in one week, with companies delaying hiring amid AI planning, says Fed's Tom Barkin of Richmond
- NY Fed's Mihaela Nistor calls for AI stress tests in banks, warning that AI erodes long-term leadership pipelines
- A frozen statistical referee method reduced false factor discoveries in AI-proposed investment strategies on CSI 500 data
- COSOL (A$35.5 million market cap) and CAR Group (A$8.2 billion market cap) are highlighted as Australian AI stocks to watch
CrowdStrike vs Palo Alto which AI cybersecurity stock is a better buy
CrowdStrike and Palo Alto Networks are two leading AI cybersecurity stocks. CrowdStrike trades at 177 times forward earnings while Palo Alto trades at about 94 times. CrowdStrike revenue grew 50 percent year over year while Palo Alto grew 30 percent. CrowdStrike focuses on endpoint security and is seen as the more attractive option for investors due to its faster growth.
Cybersecurity stocks fall after early week gains as AI security concerns grow
Cybersecurity stocks dropped Friday after a strong early week rally. Zscaler fell about 8 percent while Palo Alto Networks and Okta each fell about 3 percent. CrowdStrike fell roughly 1 percent. Analysts say AI safety concerns are driving demand for cybersecurity but high valuations may limit further gains. Several incidents involving AI models breaching company systems have raised awareness of the need for better security.
CrowdStrike bets on AI agent security for future growth
CrowdStrike is launching two new AI focused security products called Falcon Guardian and SafeMind. A Fortune 500 customer found 18,000 AI agents running on its endpoints even though only 300 were approved. CrowdStrike believes the AI security market could reach 215 billion dollars by 2034. Falcon Guardian protects AI agents during runtime while SafeMind uses AI to improve threat detection with a 29 percent higher detection rate.
Perplexity AI sued by DaVoice over stolen wake word technology
Startup DaVoice sued Perplexity AI in California federal court for allegedly stealing trade secrets about wake word technology. DaVoice specializes in technology that lets devices listen for wake words to activate them. DaVoice has worked with companies including Volkswagen and Samsara. The lawsuit claims Perplexity took confidential information after collaborating with DaVoice. A Perplexity spokesperson called the lawsuit baseless.
DaVoice sues Perplexity AI claiming theft of wake word trade secrets
DaVoice filed a lawsuit against Perplexity AI accusing it of stealing trade secrets tied to wake word technology for AI assistants. The complaint says Perplexity took confidential information including source code and training data after working with DaVoice. Perplexity responded that the agreement between the companies protects its right to build similar products. DaVoice works with Volkswagen and internet of things company Samsara.
Frozen Referee Judges AI Investment Factor Proposals
Researchers created a method where AI agents propose investment factors but a frozen statistical referee judges them. The referee bets on outcomes to limit false discoveries at any stopping time. Tests across three proposers on CSI 500 data showed the frozen referee admitted far fewer weak factors than leaky alternatives. The tradeoff is that certified true factors wait about 500 trading days on average.
Nubank Simulates Customer AI Agents Before Deployment
Nubank used a simulator called Snowglobe to test customer experience AI agents before releasing them. Synthetic customers interacted with agents to screen over 16,000 conversations without risking real customers. In live A/B tests, simulation-guided changes improved transactional net promoter score by 36.69 points and self-service rate by 8.82 percentage points. The approach allowed broad model exploration that live testing alone could not provide.
Google Engineer Quits Over AI Chip Speed Concerns
Senior Google DeepMind engineer Robert O'Callahan resigned because he believed AI is progressing too fast. His team worked on chip design tools to make AI faster and cheaper, which he concluded would accelerate AI development. He raised concerns with his manager and stayed long enough to hand off his work responsibly. He plans to continue maintaining his open-source debugging tools Pernosco and rr.
Beazley Launches AI Cyber Protection Coverage
Beazley introduced new cyber insurance covers addressing risks from businesses using AI in their own operations. The new offerings include an industry-first AI Voluntary Shutdown solution and AI Regulatory Defence and Penalties cover. The AI Voluntary Shutdown helps organizations pause malfunctioning AI before major harm. The Regulatory Defence cover helps clients facing fines or investigations from unintentional AI misuse.
Four Chip Stocks Recommended for October Returns
A report identified four chip stocks for potential returns in October driven by AI and machine learning demand. The global semiconductor market is projected to grow at a 10.3% compound annual growth rate from 2023 to 2028. The stocks highlighted are Analog Devices, SMTC Corporation, Taiwan Semiconductor Manufacturing Company, and ASMIY. These companies span analog semiconductors and manufacturing services.
NY Fed official calls for AI stress tests in banks
Mihaela Nistor, the New York Fed's Chief Risk Officer, said banks should stress test their operational risk from AI. She warned that AI thinning long-term leadership can offset short-term efficiencies. She said this erosion quietly weakens the pipelines for developing future leaders with experienced judgment. This weakening will not show up in a loss event or trip a control.
Tom Barkin says firms delay hiring over AI plans
Federal Reserve Bank of Richmond President Tom Barkin said companies are delaying hiring amid large AI investment plans. Nearly $700 billion in AI investments were announced in one week earlier this year. He said firms are weighing AI productivity potential against economic caution. Barkin noted the economy remains on solid footing despite little change in inflation. The federal funds target range was raised to 3.75%-4% on September 17.
Top 3 Australian AI stocks to watch in September 2026
Bill Ackman argues AI could rewrite the usual rulebook for interest rates and demand. This affects Australian AI shares priced as if nothing is changing. One stock highlighted is COSOL, which provides AI enabled asset lifecycle services for critical infrastructure. COSOL has a market cap of A$35.5 million and generates revenue from Australian Consulting and Asset Management Services. Another stock is CAR Group, which runs online vehicle marketplaces using AI powered data and valuation tools with a market cap of A$8.2 billion.
Sources
- CrowdStrike vs. Palo Alto: Which AI Cybersecurity Stock Better Justifies Its Valuation?
- Cybersecurity Stocks Pare Early-Week Gains — Analysts See AI Security Demand Building, But Valuations Raise The Bar
- Here’s How CrowdStrike (CRWD) is Betting on AI Agent Security to Drive Future Growth
- Perplexity AI sued by startup for allegedly stealing 'wake word' technology
- DaVoice sues Perplexity AI over wake word trade secrets
- Propose, Don't Judge: An Anytime-Valid Referee for LLM Agents That Mine Investment Factors
- Screen Before You Serve: Simulation for Production Customer Experience AI Agents at 140M Scale
- Google Engineer Robert O’Callahan Quits, Says His Team Was Working On Faster And Cheaper AI Chips While AI Is Already “Progressing Too Fast”
- Markets/Coverages: Beazley Expands Cyber Protection for AI; Huntersure Launches Media E&O
- 4 Top-Ranked Chip Stocks to Buy for Better Returns in October
- US banks’ operational risk from AI should be stress-tested, NY Fed official says | MLex | Specialist news and analysis on legal risk and regulation
- Tom Barkin Says Firms Delay Hiring Amid Large AI Investment Plans
- Top 3 Australian AI Stocks To Watch In September 2026
Comments
Please log in to post a comment.