Tech Leaders Urge Slower AI Development Amid Safety Concerns

Key Takeaways

• Tech leaders including Anthropic's Dario Amodei, OpenAI's Sam Altman, and Elon Musk urge slower AI development over safety concerns
• The S&P Kensho Cyber Security Index gained 39.6% year-to-date as cybersecurity firms like Okta, CrowdStrike, and Palo Alto Networks benefit from AI safety demand
• President Trump stated AI does not need guardrails, contrasting with growing calls for regulation
• Snowflake reported 93% year-over-year revenue growth from its AI-powered platform, with stock rising over 10%
• PepsiCo boosted throughput 20% using AI and digital twins, while BMW tested Figure 02 humanoid robots for 1,250 operating hours
• North American companies ordered 8,940 robots worth $622 million in Q2, signaling strong demand for physical AI
• Nasdaq Ventures led a Series A investment in Transient.AI to govern enterprise AI in regulated financial markets
• A Mozilla report found the performance gap between US frontier models and Chinese open-weights models narrowed to 4.4 months
• Agent-net released Webagent, an open-source tool that turns any website into a guarded AI agent with built-in safety enforcement
• Startup Harmoni raised $10 million to bring AI upgrades to roughly 280,000 U.S. factories running legacy machinery

Cybersecurity ETFs Gain as AI Safety Fears Grow

Rising concerns about AI safety are increasing demand for cybersecurity solutions. Top cybersecurity ETFs include First Trust Indxx Innovative Technology ETF (TAIT), Invesco PowerShares Cyber Security ETF (HACK), and iShares Cyber Security ETF (IBKC). TAIT rose 15% in the past year, HACK gained 12%, and IBKC climbed 10%. Experts say cybersecurity firms like Okta, CrowdStrike, and Palo Alto Networks stand to benefit as businesses seek to protect against AI-related threats.

AI Safety Concerns Drive Interest in Cybersecurity ETFs

Tech leaders including Anthropic CEO Dario Amodei, OpenAI CEO Sam Altman, and SpaceX CEO Elon Musk have called for slower AI development. Cybersecurity companies such as Okta, CrowdStrike, and Palo Alto Networks are expected to benefit from growing safety concerns. The S&P Kensho Cyber Security Index gained 6.6% this quarter and 39.6% year-to-date. Recommended ETFs include First Trust NASDAQ Cybersecurity ETF (CIBR) with $15.25 billion in assets and a 0.58% annual fee, and the Amplify Cybersecurity ETF.

Cybersecurity Stocks Rise Amid Calls for Slower AI Development

Major tech leaders are urging a more measured pace of AI development due to safety concerns. Despite potential slowdowns, cybersecurity firms like Okta, CrowdStrike, and Palo Alto Networks remain well positioned. CRCD CEO George Kurtz said security risks from powerful AI models will persist even if development slows. The S&P Kensho Cyber Security Index is up 39.6% year-to-date. Investors can gain exposure through ETFs like First Trust NASDAQ Cybersecurity ETF (CIBR) and Amplify Cybersecurity ETF.

Trump Says AI Does Not Need Guardrails

President Donald Trump stated that AI does not need guardrails and only requires a strong and smart president to serve as a sufficient safeguard against rogue artificial intelligence. The comment was made in response to growing calls from tech leaders for more regulation and safety measures in AI development. Trump's position contrasts with the increasing concern over AI safety risks.

Ossoff Claims Trump Is Compromised on AI Safety

Democratic Senate candidate Jon Ossoff accused President Donald Trump of being compromised on AI safety regulation. Ossoff cited investments made by Donald Trump Jr. as the reason for the alleged conflict of interest. The statement comes amid ongoing debate over how to regulate AI development and ensure safety measures are in place.

El Agente Potente brings agentic AI to atomistic simulations

Researchers introduced El Agente Potente, an agentic system for machine-learning interatomic potential simulations. It uses typed execution graphs to keep workflows structured and auditable. A coding agent adds flexibility for custom scientific tasks. The system was tested across materials discovery, molecular energy landscapes, adsorption, and catalysis workflows. Benchmarks confirmed its reproducibility and low LLM token cost.

Nuaware teams with HUMAN Security to fight AI-driven threats

Nuaware announced a partnership with HUMAN Security to help organizations protect digital experiences from fraud and automated attacks. The collaboration will use HUMAN's AI-powered security tools to defend digital assets. It aims to build trust as AI systems make more decisions affecting business and customers. The companies also plan joint research to create new AI-powered security solutions.

Open Chinese AI models nearly match US frontier models

A Mozilla report found the performance gap between US frontier AI models and top open-weights Chinese models has shrunk to 4.4 months. That is down from about 12 months in 2021. Many companies are switching to open models because they cut costs by up to 70%. Frontier models still lead in areas like image recognition and natural language processing. The report says the right workload should guide model choice.

Agent-net open sources Webagent for guarded AI agents

Agent-net released Webagent, an open source tool written in Go that turns any website into a guarded AI agent. It uses one JSON spec with nine slots covering retrieval, memory, guardrails, channels, and models. Every tool call passes through action.Guard before it runs, so safety is enforced in code. The system supports OpenRouter brains, MCP tools, Slack, and WhatsApp channels. It is licensed under Apache 2.0.

Nasdaq invests in Transient.AI to govern enterprise AI

Nasdaq Ventures led a Series A investment in Transient.AI, a company that governs AI use in regulated financial markets. Transient provides bank-grade controls including auditability, policy enforcement, and sandboxing for AI agents. Its Caddie.AI assistant supports the full trading lifecycle for hedge funds, banks, and asset managers. The funding will help expand operations in North America, Europe, and Asia-Pacific hubs like London, Singapore, Tokyo, and Hong Kong.

AI upgrades aging US factories without rebuilding them

PepsiCo used AI and digital twins to boost throughput 20% at an existing plant and expects to cut capital spending by 10% to 15%. BMW tested Figure 02 humanoid robots on its existing Spartanburg assembly line for about 1,250 operating hours over 10 months. Startup Harmoni raised $10 million to target roughly 280,000 U.S. factories still running legacy machinery. North American companies ordered 8,940 robots worth $622 million in the second quarter, showing growing demand for physical AI in existing plants.

Snowflake reports strong AI-driven growth in quarterly results

Snowflake reported revenue growth of 93% year-over-year, driven by its AI-powered platform. The company also saw a 25% increase in users, with many coming from the financial services sector. Its stock price rose over 10% after the earnings report. The results validated Snowflake's strategy in cloud-based data warehousing and business intelligence.

Study tests LLM judges for AI patent drafting quality

Researchers studied whether LLM judges can reliably evaluate AI-generated patent drafts through a testbed called Vibe Patenting. They found that judge-guided revision consistently improved quality, while unguided revision tended to plateau. A low-reasoning agent could match a much more expensive high-reasoning agent when using iterative judge feedback. The study also validated results against a professional patent attorney, finding meaningful but metric-dependent agreement.

Sources

• Time to Boost Cybersecurity ETF Exposure as AI Safety Concerns Rise
• Time to Boost Cybersecurity ETF Exposure as AI Safety Concerns Rise
• Time to Boost Cybersecurity ETF Exposure as AI Safety Concerns Rise
• Trump says AI doesn’t need guardrails, only a ‘STRONG AND SMART’ president
• Ossoff Says Trump Is ‘Compromised’ On AI Safety Regulation Due To Don Jr. Investments
• El Agente Potente: High-Throughput Agentic Atomistic Simulations
• Nuaware Partners with HUMAN Security to Help Organisations Build Trust in the Agentic AI Era
• Exclusive: Open Chinese models close gap with Silicon Valley’s frontier AI models
• Agent-net Open Sources Webagent: A Go Harness That Turns Any Website into a Guarded AI Agent
• Nasdaq invests in a company that limits what AI programs can do
• Physical AI’s Biggest Market Is 280,000 Aging US Factories
• Snowflake at Piper Sandler growth frontiers: AI lifts growth
• Vibe Patenting: Evaluating LLM Judges for Professional Patent-Drafting Agents