Anthropic, OpenAI Push for Stronger AI Safety Rules

Anthropic and OpenAI are pushing for stronger AI safety rules, raising concerns about how AI systems are managed and controlled. Both companies want to shape industry-wide standards around AI safety. Separately, Anthropic CEO Dario Amodei is set for a private dinner with President Trump at the White House on Sunday. Amodei has urged slowing the pace of AI model improvements, a stance that contrasts with Trump's recent dismissals of AI danger warnings from lawmakers and industry leaders.

Goldman Sachs forecasts that hyperscalers will pour $7.6 trillion into AI by 2025, far surpassing the $1.4 trillion spent on AI in 2020. Interestingly, the biggest financial winners may not be chip makers but rather companies providing software, services, and expertise tied to the AI spending boom.

Google's security team rewrote critical C libraries in Rust using Gemini AI and differential fuzzing. The project preserved original C symbols so downstream code would not break, and automated testing verified behavior across over 30 million GIF files. The rewrite uncovered a new security flaw in the original C library, while runtime performance matched the old version. The work is now open source as giflib-rust.

Alibaba unveiled the Zhenwu V900 AI chip on September 22, calling it China's most powerful AI accelerator. The chip delivers roughly three times the performance of its predecessor. Meanwhile, US Representative Jill Tokuda led 15 House Democrats in urging President Trump to avoid concessions on advanced AI chip sales to China, noting that Nvidia's Blackwell chips run up to eight times faster than older processors previously sold to China.

Okta is leaning into AI security with new capabilities announced at Oktane 2026, including support for AI agents through its Cross App Access protocol and Agent SSO features. Q2 2027 revenue rose 11% to $805 million, and RPO grew 17% to $4.86 billion. BMO Capital lifted its price objective to $230 from $187. Okta had 5,255 customers with ACV above $100,000 as of July 2026.

US businesses are increasingly adopting cheaper Chinese AI models like Baidu's ERNIE to cut costs. These open-source options are easier to integrate than systems from OpenAI or Anthropic, and the trend is spreading to Europe and Asia as the global AI market fragments. Some experts warn of risks, but companies report significant savings and improved efficiency.

Apple under new CEO John Ternus launched its first foldable smartphone, the iPhone Duo, priced at $1,999 and powered by the A20 Pro chip. The company also plans to introduce Apple Pay in India through Axis Bank credit cards and is exploring AI-focused servers designed for inference workloads using its M-series chips. Apple faces regulatory scrutiny in the EU and China that could limit product rollouts.

Key Takeaways

Anthropic and OpenAI are jointly advocating for stronger AI safety rules and industry-wide standards for managing AI systems.
Anthropic CEO Dario Amodei will meet President Trump at a private White House dinner on Sunday to discuss AI risks.
Goldman Sachs predicts hyperscalers will spend $7.6 trillion on AI by 2025, with software and services companies potentially profiting most.
Google rewrote critical C libraries in Rust using Gemini AI, finding a new security flaw in the original code while maintaining performance parity.
Alibaba launched the Zhenwu V900 AI chip, claiming it is China's most powerful AI accelerator with roughly three times the performance of its predecessor.
Okta reported 11% revenue growth to $805 million and 17% RPO growth to $4.86 billion, driven by new AI agent security features.
US companies are turning to cheaper Chinese AI models like Baidu's ERNIE, with the trend spreading to Europe and Asia.
Rep. Jill Tokuda and 15 House Democrats urged Trump to maintain strict controls on advanced AI chip sales to China, citing Nvidia Blackwell's performance advantage.
Apple launched the iPhone Duo foldable at $1,999 and is exploring AI-focused servers using M-series chips under new CEO John Ternus.
Intelligent data pipelines with self-monitoring and auto-healing capabilities are improving enterprise AI analytics for real-time applications like fraud detection.

Anthropic and OpenAI warn about AI safety risks

Anthropic and OpenAI raised concerns about AI safety. Both companies want to help control how AI systems are managed. They aim to influence safety rules for the industry.

Anthropic and OpenAI warn about AI safety risks

Anthropic and OpenAI raised concerns about AI safety. Both companies want to help control how AI systems are managed. They aim to influence safety rules for the industry.

Anthropic and OpenAI warn about AI safety risks

Anthropic and OpenAI raised concerns about AI safety. Both companies want to help control how AI systems are managed. They aim to influence safety rules for the industry.

Okta stock may shift as AI security push grows

Aembit announced support for Okta's Cross App Access protocol at Oktane 2026. This extends Okta's identity policies to AI agents. It tightens control over nonhuman access without adding extra consents for employees.

Okta AI pivot could boost growth

Okta announced new Okta for AI Agents capabilities at Oktane 2026, including Agent SSO. Q2 2027 revenue rose 11% to $805 million. RPO grew 17% to $4.86 billion. BMO Capital lifted its price objective to $230 from $187. Okta had 5,255 customers with ACV above $100,000 as of July 2026.

Google rewrites C libraries to Rust with AI and fuzzing

Google's security team rewrote critical C dependencies to Rust using Gemini AI and differential fuzzing. The team kept the original C symbols so downstream code would not break. Automated testing checked over 30 million GIF files for matching behavior. The rewrite found a new security flaw in the original C library. Runtime speed matched the old C version, and the work is now open source as giflib-rust.

Goldman Sachs predicts huge AI spending boom

Goldman Sachs says hyperscalers will spend $7.6 trillion on AI by 2025. That amount far exceeds the $1.4 trillion spent on AI in 2020. The analyst says the biggest winners may not be chip makers. Instead, companies providing software, services, and expertise could profit most from the AI spending boom.

Alibaba unveils powerful new AI chip

Alibaba Group introduced the Zhenwu V900 AI chip on September 22. The company calls it China's most powerful AI accelerator. It delivers roughly three times the performance of its predecessor. Alibaba is investing heavily in AI research and partnerships. Some experts question whether such costly AI investments can stay sustainable.

Intelligent data pipelines improve enterprise AI analytics

Traditional ETL pipelines run nightly batches and can miss errors until someone notices. Intelligent data pipelines add self-monitoring for data freshness, volume, and schema drift. They can auto-heal some failures and block bad data from reaching downstream systems. These pipelines support real-time AI products like fraud detection and LLM agents. However, smaller teams may benefit more from simpler testing tools than from full ML-based observability stacks.

Tokuda urges Trump to protect US AI chip lead

US Representative Jill N. Tokuda led 15 House Democrats in a letter to President Trump. They urged him to avoid concessions on advanced AI chip sales to China before talks with President Xi Jinping. The lawmakers invoked the Export Control Reform Act to keep strict controls in place. They noted that Nvidia's Blackwell chips run up to eight times faster than older processors sold to China.

US businesses adopt cheaper Chinese AI models

US companies are turning to Chinese AI models like Baidu's ERNIE to cut costs. These open-source models are cheaper and easier to integrate than systems from OpenAI or Anthropic. Major corporations report significant savings and improved efficiency. The trend is spreading to Europe and Asia as the global AI market fragments. Some experts warn of risks but businesses are willing to proceed.

Trump and Anthropic CEO set for White House dinner

President Trump will meet Anthropic CEO Dario Amodei for a private dinner at the White House on Sunday. Amodei raised concerns about AI risks and urged slowing the pace of AI model improvements. Trump has recently dismissed concerns from lawmakers and industry leaders about AI dangers. The meeting was confirmed by two sources familiar with the dinner. Anthropic is the company behind the Claude chatbot.

Apple eyes foldables and AI servers under new CEO

Apple launched its first foldable smartphone, the iPhone Duo, priced at $1,999 and powered by the A20 Pro chip under new CEO John Ternus. The company plans to introduce Apple Pay in India through Axis Bank credit cards. Apple is also exploring AI-focused servers designed for inference workloads using its M-series chips. The firm faces regulatory scrutiny in the EU and China that limits its product rollouts. Institutional investors like BlackRock hold significant stakes in the company.