New York City Council held a marathon hearing on AI safety on October 5, calling four major companies to estimate the risk of a catastrophic disaster. Representatives from Anthropic, Google, OpenAI, and Meta answered under oath but could not provide specific numbers. Elon Musk's SpaceXAI ignored a subpoena, and the council is now pursuing legal action to force compliance. Council Speaker Menin said she learned more from what companies refused to say than from their answers.
EPIC counsel Mayu Tobin-Miyaji also testified at the hearing, urging the council to address real harms already affecting New Yorkers. Whistleblowers from leading AI companies warned about insufficient testing before deployment. EPIC called for accountability and for preventing companies from releasing uncontrollable products while avoiding liability.
Separately, researchers uncovered a fleet of AI agents, likely tied to Chinese firm Tencent, that scraped data from rival Alibaba's Amap maps service. The agents conducted 1,810 scans in a single day, bypassed protections, and sent results to Tencent Cloud inboxes. Some runs labeled themselves as Claude, and the data appeared aimed at evaluating navigation patterns at locations like zoos and parks.
On the international front, analysts argue the United States and China could strike a meaningful AI safety deal by removing throttling on Chinese AI development in exchange for mutual pacing. The authors say both countries would benefit from slowing progress together rather than competing for dominance, since long-term suppression gives China little incentive to negotiate.
In healthcare, Toronto General Hospital surgeons are testing an AI overlay that displays green and red zones during operations to guide safe and unsafe areas. Dr. Amin Madani and Dr. Timothy Jackson used the system during a gallbladder removal and suggested adding a yellow caution zone for areas needing human judgment. Canada is seen as a leader in AI innovation, and the technology offers promise for improving care amid hospital resource shortages.
Arctic Wolf CEO Nick Schneider said AI has fundamentally changed cybersecurity, shrinking the window to respond to vulnerabilities to what he calls negative time. He believes organizations need their own AI agents to defend against sophisticated machines. Arctic Wolf has grown to about 12,000 customers, and an IPO is a question of when, not if.
Geneva-based cyber expert Stéphane Duguin warns the region is unprepared for AI-powered cyber threats targeting civil society. His organization Protect.ngo has tracked over 1,050 attacks affecting more than 900 organizations globally since 2018. A letter warns of a 6 to 12 month window before advanced models like Anthropic's Mythos become widely available, potentially overwhelming underfunded NGOs.
U.S. employers now face a growing patchwork of AI and data compliance rules. The EU AI Act treats AI tools used in hiring and performance reviews as high-risk, with transparency rules taking effect in 2026. France requires works council consultation and data impact assessments, while the UK's DUA Act shifted rules on automated decisions. Several U.S. states have also passed their own AI laws, forcing employers to coordinate across jurisdictions.
Education is another area seeing steady adoption. Daniel Huddart, an associate professor at Touro University Nevada, uses AI tools including question banks, clinical reasoning cases, and animated demonstrations to teach physical therapy students. He says the tools save time and help students grasp difficult material. Jessica Grimm, a nursing professor at the same university, stresses that AI supports teaching rather than replacing it.
Key Takeaways
- NYC Council heard testimony from Anthropic, Google, OpenAI, and Meta on AI safety, but none could quantify catastrophic risk.
- SpaceXAI ignored a subpoena to appear at the NYC Council hearing, prompting legal action.
- EPIC urged the council to hold AI companies accountable for harms already affecting New Yorkers.
- A fleet of AI agents tied to Tencent scraped Alibaba's Amap maps data with 1,810 scans in one day.
- Analysts propose a U.S.-China AI safety deal that removes throttling in exchange for mutual pacing on development.
- Toronto General Hospital surgeons are testing an AI overlay that shows safe and unsafe zones during operations.
- Arctic Wolf CEO says AI has changed cybersecurity, with the company now serving about 12,000 customers.
- Geneva faces a 6 to 12 month window before advanced models like Anthropic's Mythos become widely available, threatening underfunded NGOs.
- The EU AI Act, France's data rules, the UK's DUA Act, and various U.S. state laws are creating a compliance patchwork for employers.
- Touro University Nevada uses AI tools to teach physical therapy and nursing students, supporting rather than replacing instructors.
NYC Council Pressures AI Companies on Safety Risks
New York City Council held a marathon hearing on AI safety on October 5. Lawmakers asked four major AI companies to estimate the risk of a catastrophic disaster. None of the companies gave a specific number. Elon Musk's SpaceXAI ignored a subpoena to appear, and the council is now taking legal action to force compliance. Council Speaker Menin said she learned more from what the companies refused to say than from their answers.
EPIC Urges NYC Council to Address Real AI Harms
EPIC counsel Mayu Tobin-Miyaji testified at a New York City Council hearing on AI safety bills. Whistleblowers from leading AI companies warned about risks and insufficient testing before deployment. Representatives from Anthropic, Google, OpenAI, and Meta answered questions under oath but could not quantify catastrophic risk. EPIC urged the council to hold AI companies accountable for harms already affecting New Yorkers and to prevent companies from releasing uncontrollable products while avoiding liability.
How to Strike an AI Safety Deal with China
This article examines how the United States could approach AI safety agreements with China. The authors argue that throttling China's AI development and pacing AI progress are in tension. If the U.S. commits to long-term suppression of Chinese AI, China has less reason to make a safety deal. The authors propose that the best deal would remove throttling in exchange for mutual pacing on AI development. They argue both countries would benefit from agreeing to slow progress together rather than competing for AI dominance.
AI Helps Surgeons Navigate Safe Operations in Toronto
At Toronto General Hospital, surgeons are testing an AI system that shows safe and unsafe zones during operations. Dr. Amin Madani and Dr. Timothy Jackson used the AI overlay while removing a gallbladder. The system displayed green and red areas on a monitor to guide the surgeons. The team suggested adding a caution yellow zone for areas requiring human judgment. Canada is seen as a leader in AI innovation, and the technology offers hope to improve healthcare despite long wait times and hospital bed shortages.
We Need to Rethink What AI Really Means
This article argues that the term artificial intelligence is too vague and often tied only to large language models like ChatGPT and Claude. The author says many non-LLM AI technologies are already making breakthroughs in cancer research and fighting global warming. These systems are described as more reliable and trustworthy than chat-based AI. The article calls for expanding the definition of AI and recognizing that many forms of the technology deserve support rather than skepticism.
AI agent fleet tied to Tencent scraped Alibaba maps data
Researchers found a fleet of AI agents, likely from Chinese firm Tencent, that pulled data from rival Alibaba's Amap maps service. The fleet conducted 1,810 scans in a single day and used techniques to bypass Amap's protections. The agents sent results to inboxes created on Tencent Cloud, and some runs labeled themselves as Claude. The data may have been gathered to evaluate navigation patterns at locations like zoos and parks.
Arctic Wolf CEO says AI has changed cybersecurity forever
Arctic Wolf CEO Nick Schneider said AI has fundamentally changed cybersecurity because humans can no longer defend against sophisticated machines. He believes organizations need AI agents of their own to stay safe. He explained that the window to respond to vulnerabilities has shrunk to what he calls negative time. Arctic Wolf has grown to about 12,000 customers and Schneider says an IPO is a question of when, not if.
U.S. employers must plan for global AI and data compliance
U.S. employers can no longer focus only on domestic rules when using AI and handling employee data. The EU AI Act treats AI tools used in hiring and performance reviews as high-risk, with new transparency rules taking effect in 2026. France requires works council consultation and data impact assessments, while the UK's DUA Act shifted rules on automated decisions. Several U.S. states have passed their own AI laws, creating a patchwork that employers must map and coordinate across jurisdictions.
Geneva not ready for AI-powered cyber threats, expert warns
Geneva-based cyber expert Stéphane Duguin warns that the region is ill-equipped for AI-supercharged cyberthreats targeting civil society. His organization Protect.ngo has tracked over 1,050 attacks affecting more than 900 organizations globally since 2018. He says AI lets attackers increase the speed, scale and reach of ransomware, phishing and fraud. A letter warns of a 6 to 12 month window before advanced models like Anthropic's Mythos become widely available, which could overwhelm underfunded NGOs.
Touro professor uses AI to teach physical therapy students
Daniel Huddart, an associate professor at Touro University Nevada, uses AI to help future physical therapists understand complex concepts. He created AI tools including question banks, clinical reasoning cases and animated demonstrations, such as one showing how pain signals travel to the brain. He says the tools save time and help students grasp difficult material. Jessica Grimm, a nursing professor at the same university, stresses that AI supports teaching rather than replacing it.
Sources
- New York City Council Grills AI Giants Over Safety Risks
- EPIC to NYC Council: Address the AI Harms Already Affecting New Yorkers
- How to Make an AI Deal With China: Trade Throttling for Pacing
- Can Artificial Intelligence Learn the Art of Surgery?
- In Medias Res: ‘Behold – a man!’ and the reluctant argument for AI
- AI agent fleet likely from Chinese firm Tencent pulled data from rival Alibaba’s maps, researchers say — 1,810 scans in one day, some runs labeled themselves ‘claude’
- Arctic Wolf CEO: AI Has Changed Cybersecurity Forever
- Global AI and Data Compliance: Why U.S. Employers Can’t Afford a U.S.-Only Lens
- Cyber expert Stéphane Duguin: International Geneva ill-equipped for AI-supercharged cyberthreats
- Touro University Nevada professor uses AI to teach future physical therapists
Comments
Please log in to post a comment.