Anthropic reports that Claude now leads 26% of its internal AI research, a sharp rise from just 1% earlier this year. The company tracks a Claude research agent that reviewed roughly 15,000 R&D tasks across 542 categories. Over 90% of Anthropic's research now involves some form of human-AI collaboration, and the firm plans to publish these metrics regularly with independent evaluators.
Meanwhile, Bart Smith, CEO of Avalanche Treasury Company, warns that AI agents trading in financial markets could strain blockchain capacity. Because AI agents operate without breaks, transaction activity stays constant around the clock. Smith expects traditional finance to move toward 24/7 operations by mid-2027, and Avalanche is upgrading its network with ACP-176 and ACP-194 to handle the growing load.
On the policy front, the Defense Production Act, a Cold War-era tool from 1950, could be used by President Trump against dangerous AI systems. Its authorities were extended through Dec. 11, 2026. AI company leaders have recently warned about increasingly capable models and called for slower development to let safety measures catch up. Separately, the EFF urges lawmakers to ground AI cybersecurity rules in proven best practices like stronger sandboxing and monitoring, and to require public third-party investigations into serious security incidents.
A Pew Research Center survey of 50,838 people across 37 countries found that more people expect AI to eliminate jobs than create them. In the U.S., 71% of respondents predicted fewer jobs due to AI, compared with a global average of 46%. Wealthier countries showed more concern about job loss. At Harvard, faculty resistance is growing as the university pushes to adopt AI in teaching, with educators worried about the replacement of human instructors and the impact on low-income students.
Key Takeaways
- Anthropic says Claude now leads 26% of its AI research, up from 1% earlier this year
- A Claude research agent reviewed about 15,000 R&D tasks across 542 categories
- Bart Smith warns AI agent trading could strain blockchain capacity, pushing finance toward 24/7 operations by mid-2027
- The Defense Production Act authorities were extended through Dec. 11, 2026, and could be used against dangerous AI systems
- Pew Research finds 71% of U.S. respondents expect AI to cause job loss, compared with a 46% global average
- The EFF calls for AI cybersecurity rules based on proven best practices and public third-party investigations
- Harvard faces faculty resistance as it pushes AI adoption in classrooms
- Huawei plans to release Ascend 960DT in Q1 2027, Ascend 960PR in Q3 2027, and Ascend 970 and 980 in 2028 and 2029
- Anthropic plans to publish its AI usage metrics regularly and bring in independent evaluators
- Passing AI safety legislation this year is considered unlikely despite ongoing congressional debate
Anthropic says Claude now handles 26% of its AI research
Anthropic reported that its AI system Claude now leads 26% of its research and development work. This is up from just 1% in March. Over 90% of the company's research now involves human-AI collaboration. Anthropic created a new index to measure how much AI is used in its development work.
Claude leads 26% of AI research at Anthropic
Anthropic published new internal metrics showing Claude leads 26% of its AI research as of August 2026. Over 90% of work involves AI collaboration. A Claude research agent reviewed about 15,000 R&D tasks across 542 categories. Anthropic plans to publish these metrics regularly and bring in independent evaluators.
AI agent trading could strain blockchain capacity
Bart Smith, CEO of Avalanche Treasury Company, warned that AI agents trading in financial markets could exhaust blockchain block space. AI agents work without breaks, creating constant transaction activity. Smith predicted traditional finance may shift to 24/7 operations by mid-2027. Avalanche is upgrading its network with ACP-176 and ACP-194 to handle the expected load.
Survey finds people fear AI job loss more than job growth
A Pew Research Center study found more people expect AI to cause job loss than create jobs over 20 years. In the U.S., 71% of respondents predicted fewer jobs due to AI. The global average was 46%. The study surveyed 50,838 people across 37 countries between February and June. Wealthier countries showed more concern about job loss.
Cold War-era tool for AI emergency set to expire in December
The Defense Production Act, a Cold War-era tool from 1950, could be used by President Trump against dangerous AI systems. Its authorities were extended through Dec. 11, 2026. The act covers cyberattacks, exports, federal contracts and civil liability. AI company leaders recently warned about increasingly capable models and called for slower development to allow safety measures to catch up.
AI kill switch could shut down the internet
A 'kill switch' for AI is a guardrail that could stop AI agents from taking unwanted action. At a small scale it could be as simple as pulling a plug. On a global level it might go as far as shutting down the internet. Lawmakers and industry leaders are debating how to implement these safety measures.
EFF urges lawmakers to ground AI cybersecurity rules in best practices
The EFF told lawmakers that new AI cybersecurity rules should be based on proven security best practices. Following practices like stronger sandboxing and monitoring could have prevented security breaches at AI labs. The group also wants independent third-party investigations into serious security incidents. These reports should be made public to ensure oversight.
Huawei reveals new AI accelerator roadmap and Ascend NPUs
Huawei updated its AI hardware roadmap with new accelerators and pulled in next-generation Ascend 960 processors. The Ascend 960DT is set for Q1 2027 and the Ascend 960PR follows in Q3 2027. Huawei plans to release the Ascend 970 and 980 chips in 2028 and 2029. Production capacity remains insufficient to meet domestic demand.
Harvard AI debate heats up as faculty resist classroom adoption
Harvard is pushing to adopt AI in teaching and operations but faces growing resistance from faculty. A Harvard Business Review report found AI is being used in more classrooms. Many educators worry AI could replace human teachers and undermine liberal arts education. There are also concerns about AI worsening inequalities for low-income students.
Key lawmakers to watch on AI policy in Congress
Lawmakers are debating how to regulate AI after dire warnings from AI researchers. Ideas floated include new caucuses, bills, and classified briefings. Passing AI safety legislation this year is considered unlikely. Rep. Jay Obernolte of California is a key voice due to his engineering and AI background.
Sources
- AI's role in building AI surging? Anthropic says Claude now leads 26% of its R&D
- Claude Is Leading 26% Of AI Research At Anthropic, Collaborating On Over 90%
- AI Agent Trading Surge Could Push Blockchain Capacity to Its Limits
- People fear AI will cause job loss, not growth, new survey finds
- Powerful Cold War-era tool Trump could use in an AI emergency is set to expire in December
- An AI 'kill switch' could go as far as shutting down the internet
- EFF to Lawmakers: Ground AI Cybersecurity Rules in Best Practices
- Huawei details AI accelerator roadmap, pulls in next-generation Ascend NPUs by several quarters — FP4 performance of the Ascend 960PR doubles expectations
- The Harvard AI Debate Heats Up
- Who to watch in Congress on AI policy
Comments
Please log in to post a comment.