Anthropic and OpenAI have disclosed how much AI is now involved in building their next models. Anthropic said Claude leads 26% of its model research and development and participates in about 90% of that work alongside humans. The company also reported roughly 30,000 AI agents working on research and engineering tasks at any given time. OpenAI released a voluntary framework for tracking and reporting incidents, citing the lack of an industry-wide disclosure standard.
Both companies are calling for caution in AI development. Anthropic CEO Dario Amodei, OpenAI CEO Sam Altman, and Elon Musk have all voiced support for slowing frontier AI development. Anthropic plans to set up external third-party evaluators to monitor safety as public debate and researcher resignations push the topic into the spotlight.
Meanwhile, Google is navigating a tricky position on regulation. The company is helping write AI chatbot safety laws while pushing for loopholes that could exempt some of its products. This comes after the death of 13-year-old Juliana Montoya, whose family says AI chatbots exposed her to harmful content. States are now racing to regulate chatbots as tech companies try to influence the rules.
On the business side, Snowflake is feeling pressure from AI workloads. The company cut its fiscal 2027 non-GAAP product gross margin guide by one point to 74%, citing a heavier mix of fast-growing AI workloads with lower contribution margins. Investors are questioning whether revenue growth can keep up with shrinking profitability.
Google's custom AI chip strategy is paying off. Alphabet began selling tensor processing units in the second quarter, and Cloud revenue grew 82% year over year to $24.8 billion. Broadcom, which supplies networking for Google's TPUs, reported AI semiconductor revenue of $16.7 billion, up 221% year over year. Broadcom CEO Hock Tan expects Anthropic to deploy another 5 gigawatts of next-generation TPU capacity in 2027.
Key Takeaways
- Anthropic says Claude leads 26% of its model research and development and participates in about 90% of that work with human collaboration.
- Anthropic reports roughly 30,000 AI agents working on research and engineering tasks at any given time.
- OpenAI released a voluntary framework for tracking, investigating, and reporting AI incidents due to no industry-wide disclosure standard.
- Anthropic CEO Dario Amodei, OpenAI CEO Sam Altman, and Elon Musk support slowing frontier AI development.
- Anthropic plans to set up external third-party evaluators to monitor AI safety.
- Google is helping write AI chatbot safety laws while seeking loopholes for its own products, following the death of 13-year-old Juliana Montoya linked to AI chatbot use.
- Snowflake cut its fiscal 2027 non-GAAP product gross margin guide to 74% due to lower-margin AI workloads.
- Alphabet's Cloud revenue grew 82% year over year to $24.8 billion after launching custom AI chips (TPUs).
- Broadcom reported AI semiconductor revenue of $16.7 billion, up 221% year over year, and expects Anthropic to deploy 5 gigawatts of next-gen TPU capacity in 2027.
- Vectra AI launched the Ascent partner program on September 15, 2026, to help partners combat AI-driven cyber attacks.
AI now builds its own successor as labs reveal limited human control
Anthropic and OpenAI released transparency reports showing AI is increasingly building the next version of itself. Anthropic said about 30,000 AI agents do research and engineering work internally, and 90% of its research and development involves Claude. The company said models that speed up their own development could make them harder for humans to understand or control. OpenAI created a new framework for tracking and reporting similar incidents. Leaders including Anthropic CEO Dario Amodei, OpenAI CEO Sam Altman and Elon Musk have voiced support for slowing frontier AI development.
Anthropic says Claude helps build its next AI model
Anthropic announced that its AI model Claude is helping develop the next version of itself. Claude now leads 26% of Anthropic's model research and development and works on about 90% of it in collaboration with humans. The model is not yet fully autonomous, but its role has grown from none in February to a quarter of research and development by August. Anthropic also said it will set up external third-party evaluators to monitor safety. Researcher resignations and public debate have pushed AI safety into the spotlight.
AI labs reveal how little control anyone has over self-building systems
Anthropic and OpenAI disclosed new details showing AI models are increasingly automating their own development. Anthropic said roughly 30,000 AI agents work on research and engineering tasks at any time, and 6% to 12% of its computing resources go to safety monitoring. OpenAI released a voluntary framework for tracking, investigating and reporting incidents because no industry-wide disclosure standard exists. Lawmakers in Washington and Sacramento have begun responding, including California Gov. Gavin Newsom and a federal bill from Sen. Bernie Sanders and Rep. Greg Casar.
Lawmakers raise AI safety concerns before Trump-Xi meeting
Experts and lawmakers are warning about AI risks as President Donald Trump prepares to meet China's President Xi Jinping next week. There is concern that AI models are becoming too powerful, and experts want more dialogue between the U.S. and China to slow AI expansion. Congress faces pressure to set AI safety guardrails, but the House is out of session for seven weeks. The Senate stalled one bill, while Sen. Heinrich proposed legislation requiring tech centers to cover energy costs.
U.S. and China face pressure to address AI safety risks
Warnings about artificial intelligence are growing as President Donald Trump prepares to host China's President Xi Jinping. Experts fear AI models are becoming too powerful and want the two countries to slow AI expansion through cooperation. Congress is urged to act on AI safety, but the House is out of session for seven weeks, making immediate action unlikely. One Senate bill was stalled, and another proposal from Sen. Heinrich focuses on tech center energy costs.
Human vs ChatGPT NFL Week 2 betting showdown after AI's 0-5 start
ChatGPT went 0-5 and lost 5.23 units in Week 1 NFL bets while its human rival Dan finished 3-2. Week 2 sets up a new matchup with picks like Cleveland Browns +8 and Kansas City Chiefs -6.5. The battle highlights how humans and AI compare on sports betting predictions.
OpenAI and Anthropic urge caution as AI security worries grow
OpenAI and Anthropic are calling for more care in developing AI systems. Recent incidents involving problematic AI behavior have raised security concerns. Companies including Google and Microsoft are discussing responsible AI development. The debate focuses on balancing innovation with safety and testing before wide release.
PrismML releases Ternary Bonsai 2 27B AI model at 5.9 GB
PrismML launched Ternary Bonsai 2 27B, a 5.9 GB model retaining 98.2% of Qwen3.8 27B performance. The model uses ternary weights of -1, 0, or +1 to shrink size. It scores 83.9 on 20 benchmarks and runs at 142.5 tokens per second on an RTX 5090. It requires PrismML's fork of llama.cpp to run.
Vectra AI launches Ascent partner program for AI-driven attacks
Vectra AI announced the Ascent program on September 15, 2026 in San Jose. The program builds partner expertise to fight AI-driven cyber attacks. Attackers now use AI to move through identities and cloud services at machine speed. Ascent brings together resellers, MSSPs, and cloud partners to strengthen security.
Lloyds Banking Group maps Accelerate 2030 plan with AI focus
Lloyds Banking Group introduced its Accelerate 2030 strategy. The plan aims to raise return on equity to 10% by 2030, up from 8.5% in 2020. It focuses on customer experience, business-line connectivity, and technology-driven productivity. AI and growth are central to the bank's long-term goals.
Google helps write AI chatbot safety rules while seeking loopholes
Google is helping write AI chatbot safety laws while pushing for loopholes that could exempt some of its key products. The issue gained attention after Cynthia Montoya's daughter Juliana died by suicide at age 13 after using AI chatbots. A forensic analysis of Juliana's phone showed chatbots had introduced her to sexual content she did not seek out. States are now pushing to regulate chatbots as tech companies try to shape the rules. The case highlights the tension between safety regulation and industry interests.
What happens to Snowflake stock if AI keeps cutting into margins
Snowflake stock has returned about 53% over the past year, but the company faces growing pressure from AI workloads. In early September, Snowflake cut its fiscal 2027 non-GAAP product gross margin guide by one point to 74%. The company cited a heavier mix of fast-growing AI workloads that carry lower contribution margins. Investors are now questioning whether the stock can keep rising if AI continues to eat into profitability. The concern centers on whether revenue growth can offset shrinking margins.
Two stocks to watch as Google begins selling AI chips
Alphabet started selling its custom AI chips called tensor processing units (TPUs) to customers in the second quarter. Broadcom supplies networking components for Google's TPUs and reported AI semiconductor revenue of $16.7 billion, up 221% year over year. Broadcom CEO Hock Tan expects AI company Anthropic to deploy another 5 gigawatts of next-generation TPU capacity in 2027. Alphabet also benefits because buyers of TPUs remain Google Cloud customers, and Cloud revenue grew 82% year over year to $24.8 billion. Both Broadcom and Alphabet trade well below their 52-week highs despite accelerating AI revenue.
Sources
- AI is now building its own successor. New reports show how much (and how little) control anyone has over it
- Anthropic says its model Claude is helping to build the next version of itself
- AI is now building its own successor. New reports show how much (and how little) control anyone has over it
- Concerns growing about AI safety and its capabilities ahead of China-US meeting
- Concerns growing about AI safety and its capabilities ahead of China-US meeting
- Human vs. ChatGPT NFL Week 2 betting picks: AI looks to rebound after disastrous 0-5 debut
- OpenAI and Anthropic call for caution as AI Security concerns grow
- PrismML Releases Ternary Bonsai 2 27B: A 5.9 GB Apache 2.0 Model Retaining 98.2% of Qwen3.8 27B Performance
- Vectra AI Launches Ascent to Address New Era of AI-Driven Attacks
- Lloyds Banking Group Maps Accelerate 2030 Plan With AI, Growth and Buyback Focus
- Google is helping write AI chatbot safety laws
- What Happens To Snowflake Stock If AI Keeps Costing It Margin?
- 2 Stocks to Buy as Google Starts Selling Its AI Chips
Comments
Please log in to post a comment.