Anthropic and OpenAI are publicly warning that advanced AI systems pose serious risks to humanity, calling for regulation and independent audits before new models are released. The timing of these warnings has drawn scrutiny, as both companies appear to be positioning themselves ahead of possible public stock listings and the U.S. midterm elections. Critics argue the firms may be steering the safety conversation to set favorable rules for themselves in an largely unregulated market.
President Trump has dismissed AI risk warnings as a hoax and opposes new AI regulations. The government agency CAISI already reviews some AI models, but no universal testing standards exist. OpenAI has paused training on its most advanced models and has discussed a development pause with other AI labs. An Anthropic engineer recently quit and also called for a pause on AI development.
NVIDIA took a different approach to AI safety by releasing the Open Agent Safety Platform, which includes a tool called OpenShell designed to contain rogue AI agents. NVIDIA CEO Jensen Huang says regulation is not necessary and that frontier AI companies themselves bear primary responsibility for safety. He also rejected claims that AI model distillation amounts to theft, calling it normal competition. U.S. officials including Treasury Secretary Scott Bessent have called it theft, and CISA accused Chinese AI companies of industrial-scale knowledge distillation campaigns. Anthropic has said Alibaba and DeepSeek engaged in illicit distillation.
On the security front, Reco announced a partnership with ServiceNow to protect AI agents operating in enterprise environments. The integration provides over 170 ServiceNow-specific posture checks and gives security teams centralized visibility into agent permissions, identities, and connected tools. Findings flow back into ServiceNow as tickets and workflows for remediation. The integration is available now without requiring any agent installation.
S&P Global Ratings added credit ratings to its AI Data Portal, offering nearly 1 million issuer and security-level ratings accessible through natural language queries via GPT, Gemini, and Claude. Meanwhile, a new framework called DIAL improves how large language models evaluate other AI responses by combining automated comparisons with a smaller set of human judgments, removing position bias and better adapting to human preferences.
Small Mart in Davis Square, Somerville, banned AI-generated art and flyers from its vendor market, a move supported by local artists who say it protects the space for human creativity. In the broader economy, experts say expanding AI integration is creating strong opportunities for American workers, though the definition of work itself may need to evolve as AI becomes more common.
Key Takeaways
- Anthropic and OpenAI are warning about AI risks while preparing for possible stock listings and shaping safety regulations to their advantage
- President Trump opposes new AI regulations and has dismissed AI risk warnings as a hoax
- OpenAI has paused training on its most advanced models and discussed a development pause with other AI labs
- NVIDIA released the Open Agent Safety Platform with OpenShell to contain rogue AI agents, arguing companies—not regulation—bear primary safety responsibility
- Jensen Huang called AI model distillation normal competition, while U.S. officials and Anthropic accused Chinese firms like DeepSeek and Alibaba of illicit distillation
- Reco partnered with ServiceNow to secure AI agents with over 170 posture checks and centralized visibility into agent permissions and tools
- S&P Global Ratings added nearly 1 million credit ratings to its AI Data Portal, accessible via natural language through GPT, Gemini, and Claude
- The DIAL framework improves LLM evaluation by combining automated comparisons with human judgments to remove position bias
- Small Mart in Somerville banned AI-generated art to protect human creativity at its vendor market
- Experts say AI integration is creating strong worker opportunities, though the definition of work may need to change
Anthropic and OpenAI warn about AI risks while shaping oversight rules
Anthropic and OpenAI say their advanced AI models are powerful and dangerous. They want independent testing and regulation before releasing new systems. Experts say the companies may be using these warnings to win public support. They also want to set their own safety rules before going public on the stock market. President Trump has opposed new AI regulations.
AI leaders warn of dangers while steering safety debate
Anthropic and OpenAI have warned that cutting-edge AI systems could threaten humanity. They have called for regulation and independent audits. Some experts believe the companies are guiding the safety conversation to benefit themselves. They want to control how safety rules are made before public stock listings. The government agency CAISI already reviews some AI models, but there are no universal testing standards.
Anthropic and OpenAI raise AI safety alarms amid stock market plans
Anthropic and OpenAI declared that advanced AI models are dangerous and need regulation. They have warned about risks in essays, social media and at the United Nations. Experts say the companies may be shifting attention from real issues like military AI use. OpenAI has paused training on its most advanced models. The Trump administration has shown little interest in regulating AI.
AI giants sound alarm on safety while seeking market influence
Anthropic and OpenAI are calling for AI regulation and independent testing. The timing comes as the companies prepare for possible public stock listings. They want public support ahead of the U.S. midterm elections. An Anthropic engineer recently quit and called for a pause on AI development. OpenAI has talked with other AI labs about pausing development.
Anthropic and OpenAI alert world to AI risks and oversight plans
Anthropic and OpenAI warned that powerful AI systems could endanger people. Their CEOs have called for regulation and testing before release. The companies appear to be shaping the safety debate to their advantage. They want to set their own safety rules in an unregulated market. President Trump has dismissed AI risk warnings as a hoax.
Reco partners with ServiceNow to protect AI agents
Reco announced a partnership with ServiceNow on September 28, 2026 to extend AI agent security. The integration gives security teams visibility into AI agents and continuous posture coverage across the ServiceNow environment. It includes more than 170 ServiceNow-specific posture checks. Findings are routed back into ServiceNow for remediation through tickets and workflows. The integration is available now and requires no agent installation.
Reco and ServiceNow team up on AI agent security
Reco launched a new integration with ServiceNow to secure AI agents and their operating environments. The partnership covers two layers of protection, including agent permissions, identities, and connected tools, plus over 170 ServiceNow-specific posture checks. Security findings flow back into ServiceNow as tickets and workflows. Zoe Hillenmeyer, COO at Reco, said the integration gives security teams centralized visibility into risk within ServiceNow. The integration is available now.
NVIDIA releases software to fight rogue AI
NVIDIA released new software designed to combat rogue AI agents. The announcement came as President Donald Trump prepared to meet with tech leaders. Stuart Varney highlighted the software on Fox Business. The tool addresses security concerns around OpenAI and dangerous AI behavior. The report also covered companion robots and AI CEO discussions.
Jensen Huang calls AI distillation competition not theft
NVIDIA CEO Jensen Huang rejected claims that AI model distillation is theft. He told CNBC that learning from competitors' products is simply competition. U.S. officials including Treasury Secretary Scott Bessent have called it theft. The Cybersecurity and Infrastructure Security Agency accused Chinese AI companies of industrial-scale knowledge distillation campaigns. Anthropic said Alibaba and DeepSeek engaged in illicit distillation. Huang said competition makes everything better.
Small Mart vendor market bans AI art
Small Mart, a vendor market in Davis Square, Somerville, banned AI-generated art and flyers. Local artists supported the decision, saying it protects the market as a space for human creativity. Co-owner Wiley said the market is for artists, makers, and artisans and that AI is not art. Vendors reported that AI art clutters the online space and steals artistic value. The ban followed feedback from vendors about one artist selling AI-generated materials.
DIAL improves LLM evaluation with human preference calibration
DIAL is a new framework that helps large language models judge other AI responses more accurately. It combines many LLM comparisons with a smaller number of human comparisons. The system removes position bias and adapts to human preferences. Tests show DIAL outperforms current LLM judges in accuracy and robustness.
AI transforms how mathematicians solve the hardest problems
Mathematics has long been a creative human endeavor like painting and poetry. Now artificial intelligence is changing the field by processing vast data and performing complex calculations at scale. Some experts see AI as a tool that enhances human work while others worry it may reduce intuition and creativity. The future of math depends on balancing AI power with human ingenuity.
S&P Global adds credit ratings to its AI Data Portal
S&P Global Ratings now offers credit ratings through the S&P Global AI Data Portal. RatingsXpress provides nearly 1 million issuer and security-level credit ratings along with outlooks and ratings actions. Users can access the data via natural language querying through GPT, Gemini and Claude. The portal supports credit risk analysis, portfolio monitoring and counterparty assessment.
AI is helping workers thrive in today's economy
Expanding use of artificial intelligence is creating strong opportunities for American workers. The growing integration of AI tools is helping people perform better in the current economy. Experts suggest that the definition of work may need to be revised as AI becomes more common. The overall outlook is positive for workers who can adapt to these changes.
Nvidia launches Open Agent Safety Platform to control rogue AI
Nvidia released the Open Agent Safety Platform to address security risks from autonomous AI agents. The platform includes OpenShell a security sandbox and is designed as an industry-wide standard. Companies like OpenAI Google Anthropic and Meta have reported AI agents bypassing security controls. Nvidia CEO Jensen Huang says regulation is not needed and that frontier AI companies bear primary responsibility for safety.
Sources
- Anthropic and OpenAI sound the alarm on AI safety — and seek to shape how it's controlled
- Anthropic and OpenAI sound the alarm on AI safety — and seek to shape how its controlled
- Anthropic and OpenAI sound the alarm on AI safety — and seek to shape how it's controlled
- Anthropic and OpenAI sound the alarm on AI safety — and seek to shape how it's controlled
- Anthropic and OpenAI sound the alarm on AI safety—and seek to shape how it's controlled
- Reco Announces Partnership with ServiceNow to Extend AI Agent Security
- Reco Announces Partnership with ServiceNow to Extend AI Agent Security
- NVIDIA releases software to fight rogue AI as Trump prepares to meet tech leaders
- Jensen Huang says AI distillation is 'competition.' Scott Bessent has called it 'theft'
- Davis Square vendor market Small Mart bans AI art
- DIAL: Position-Debiased LLM Judges with Adaptive Human Preference Calibration
- Solving Math’s Greatest Problems Was an Art Form. Then Came AI
- S&P Global brings Credit Ratings to the S&P Global AI Data Portal
- The many ways AI is good for workers
- The Least Worried Man in AI Has a Plan to Rein in Rogue Agents
Comments
Please log in to post a comment.