OpenAI's Infrastructure Lead Departs; Computing Power Spending Up to $750B

Key Takeaways

• 90% of security leaders say AI attacks have changed their jobs more than anything in a decade, per a Chainguard survey.
• 44% of organizations expect to fight AI cyberattacks with AI, according to a Palo Alto Networks survey.
• 29% of vulnerabilities in 2026 were exploited in under 24 hours, far faster than the 55-day industry average patch time.
• OpenAI infrastructure lead Sachin Katti departed after roughly a year; Uday Ruddarraju now leads both compute and infrastructure units.
• OpenAI raised its computing power spending estimate to about $750 billion through 2030, up from $600 billion earlier this year.
• Anthropic warned that AI worms could self-replicate and take over smart home devices like ovens and refrigerators without human action.
• Billionaires Sunil Bharti Mittal and Noubar Afeyan called for AI regulation at the Bloomberg New Economy Forum.
• mini-Lean-agent solved all 12 Putnam 2025 problems using Astra at an average cost of $1.28 per problem.
• Mozilla released Firefox 158 with AI-powered form filling and fixes for 77 security flaws, including 30 high-risk ones.
• The ASA AI Impact Index uses Anthropic and O*NET data to track how AI affects specific job tasks across occupations.

AI Attacks Outpace Security Teams, Chainguard Survey Finds

A survey of 1,300 security leaders by Chainguard found that 90% say AI attacks have changed their job more than anything in a decade. 77% expect an attacker to exploit a vulnerability in their organization within the next year. 54% say their companies knowingly leave high or critical vulnerabilities unpatched at least sometimes. 64% report that concern about AI-accelerated attacks contributed to their burnout. The report calls for security and engineering teams to work together on prevention rather than just patching faster.

AI Transforms Critical Infrastructure Security, Survey Shows

A Sapio Research survey for Palo Alto Networks finds that 44% of respondents expect to fight AI cyberattacks with AI. The industry average patch deployment time is 55 days, while 29% of vulnerabilities in 2026 were exploited in under 24 hours. 42% say legacy OT assets that cannot be patched are the biggest cybersecurity risk. 60% reported a significant security breach in the past year. Only 32% have complete real-time visibility into all assets connected to their operational technology networks.

OpenAI Infrastructure Lead Sachin Katti Departs

OpenAI infrastructure lead Sachin Katti announced he is leaving his role after roughly a year at the company. He will remain as an advisor until November 15. Uday Ruddarraju, OpenAI's CTO of compute, will replace him and lead the company's overall compute organization. Katti said he decided to move on to build something new. He expressed pride in the team's accomplishments securing double-digit gigawatts of new compute capacity across multiple projects.

OpenAI Expands Uday Ruddarraju's Role to Lead Compute

OpenAI appointed Uday Ruddarraju as head of both compute and infrastructure units. Ruddarraju joined OpenAI in 2025 and became CTO for compute in July 2026. He previously served as head of infrastructure engineering at AWS. OpenAI expects to spend about $750 billion on computing power through 2030, up from $600 billion reported earlier this year. The company is targeting roughly $1 billion in revenue this year.

mini-Lean-agent Shows Coding Agents Can Prove Theorems

Researchers introduced mini-Lean-agent, a minimal agent for Lean theorem proving whose only model-visible tool is bash. Experiments with Gemini 3.8 Flash showed it solved 7 of 12 Putnam 2025 problems at a $2 per problem limit. With Astra as the base model, it solved all 12 problems at an average cost of $1.28 per problem. The results suggest a generic coding agent with a minimal harness can achieve strong performance on Lean benchmarks.

AI industry faces massive revenue and copyright challenges

The AI industry has multiple big problems that are getting worse at the same time. Companies need to earn around $6 trillion in revenue to match the car industry. Major AI labs face copyright crises because their technology was built using stolen content. AI data centers are also polluting and destroying infrastructure. Despite all these issues, companies focus mainly on hitting unrealistic profit targets.

Billionaires urge AI rules as Anthropic warns of self-replicating worms

Billionaires Sunil Bharti Mittal and Noubar Afeyan called for AI regulation at the Bloomberg New Economy Forum in New Delhi. Anthropic warned that AI worms could invade smart home devices like ovens, refrigerators, and televisions without human action. The company said some AI models may be months away from self-replicating and taking control of those devices. OpenAI and other leaders also stressed safety over fast growth.

Paul Weiss and Columbia Law School launch Gen AI initiative

Columbia Law School and the law firm Paul Weiss launched a five-year initiative to teach responsible generative AI use in legal education. The program will include new curricula, research, and a scholar-in-residence who will connect academics, judges, policymakers, and industry leaders. Paul Weiss has already been using AI tools and training across the firm.

Firefox 158 brings AI form filling and fixes 77 security flaws

Mozilla released Firefox 158 for Windows, macOS, Linux, and Android. The update adds an AI-powered Smart Windows feature for automatic form filling and support for the JPEG XL image format. It also fixes 77 security flaws, including 30 high-risk vulnerabilities. Users are told to update right away to close dangerous gaps. Firefox 159 is set for release on October 27, 2026.

New ASA AI Impact Index shows which jobs AI affects most

The American Staffing Association released a new tool called the ASA AI Impact Index. It uses data from Anthropic and the U.S. Department of Labor's O*NET to show how AI affects tasks and jobs. The index measures which parts of occupations are automated or changed by AI. Results give a snapshot of AI's growing impact on the labor market, and the index will be updated regularly.

Jim Cramer discusses AI power plays on Mad Money

Mad Money host Jim Cramer talked about technical analysis of AI power players. The segment focused on stocks and companies leading the artificial intelligence industry. Cramer shared his insights on which AI-related stocks are performing well. The discussion highlighted the growing importance of AI in the stock market.

PolyGraph improves cleanup of AI knowledge graphs

Agents build knowledge graphs as memory, but keeping them accurate is difficult. When sources are retracted, old information can spread through entity merges and summaries. PolyGraph asks constructors for receipts that record which evidence supported each artifact. Cleanup then uses simple set arithmetic instead of running new model calls. Tests showed that one receipt per sentence achieved 100 and 88 percent faithfulness and reduced contaminated queries to under 1 percent.

Sources

• Chainguard Research: Nine in Ten Security Leaders Say AI Attacks Have Changed Their Job More Than Anything in a Decade, but the Ability to Act Hasn't Kept Pace
• Survey Sees AI Transforming Critical Infrastructure Security for Better and Worse
• OpenAI’s Sachin Katti to leave infrastructure role, stay on as advisor
• OpenAI Expands Uday Ruddarraju’s Role to Head Compute and Infrastructure Units
• Do Theorem Provers Need Theorem-Proving Agents? mini-Lean-agent: A Coding-Agent Baseline for Lean
• The Big Problem Every Artificial Intelligence Company Faces
• Billionaires call for AI regulation as Anthropic warns of self-replicating worm threat
• Paul Weiss, Columbia Law School Launch Collaborative Gen AI Initiative
• Firefox 158 adds AI form filling and fixes 77 security flaws
• New ASA AI Impact Index Reveals the Jobs Most Impacted by AI
• Jim Cramer goes off the charts with AI power plays
• Keep Your Receipts: Trustworthy Cleanup of Agent-Built Knowledge Graphs