OpenAI President Greg Brockman says the company has slowed its most advanced AI work because of safety concerns. The move reflects a broader debate over how fast AI should develop. Treasury Secretary Scott Bessent weighed in soon after, telling Congress that AI labs should not receive liability exemptions. His remarks came days after Anthropic's Dario Amodei called for voluntary safety standards. Bessent also urged support for open-source AI built in the U.S. to stay competitive with China.
China pushed back against calls to slow AI development, with foreign spokesperson Guo Jiakun warning that fearmongering could hurt global governance efforts. Separately, security threats are growing: a pro-Ukraine group called Hacking Cat has stepped up cyberattacks on Russia, and a mass-scanning campaign is targeting exposed Vite development servers to steal cloud credentials. Security experts advise updating software and rotating secrets to reduce exposure.
On the business side, Salesforce is making moves in AI. The company launched the beta of Salesforce in Claude, a plugin built with Anthropic that brings pipeline data, accounts, and opportunities into Claude. About 7,000 Salesforce sellers are already using it, and companies like GitLab, Siemens, and Legora have deployed the tool. Salesforce also used its Dreamforce conference to highlight its AI strategy, and its stock rose after a strong earnings report.
Meanwhile, a Nationwide survey found that businesses are adopting AI faster than they manage its risks. Sixty percent of business owners say employees use public chatbots for work, but only 36% have written policies. About 31% of small and mid-sized companies report being hit by a generative AI scam in the past year. Most say they need more help defending against AI-enabled attacks.
Job seekers are also pushing back. Roughly two-thirds of 1,200 people surveyed walked away from a hiring process because of an AI interview. Candidates described the experience as cold and impersonal. The ACLU has raised concerns about bias and discrimination in these automated systems.
Key Takeaways
- <ul><li>OpenAI President Greg Brockman says safety concerns have slowed advanced AI development at the company.</li><li>Treasury Secretary Scott Bessent says AI labs should not get liability exemptions and urges more U.S.-built open-source AI.</li><li>China rejects calls to slow AI development, warning that fearmongering undermines global governance.</li><li>Hacking Cat escalates cyberattacks on Russia; mass scans target exposed Vite servers to steal cloud credentials.</li><li>Salesforce and Anthropic launch Salesforce in Claude beta, giving 7,000 sellers access to pipeline data and account summaries inside Claude.</li><li>GitLab, Siemens, and Legora are early adopters of the Salesforce in Claude plugin with 37 sales skills.</li><li>Nationwide survey: 60% of business owners say employees use public AI chatbots, but only 36% have written AI policies.</li><li>31% of small and mid-market businesses report being targeted by a generative AI scam in the past 12 months.</li><li>Two-thirds of 1,200 job seekers quit AI-driven interviews, citing cold, repetitive, and insulting experiences.</li><li>The ACLU raises concerns about bias and discrimination in AI hiring tools.</li></ul>
OpenAI president says safety concerns slowed advanced AI work
OpenAI President Greg Brockman says the company has slowed cutting-edge AI development due to safety concerns. The statement comes as AI capabilities continue to advance rapidly. Brockman emphasized that safety plays a key role in guiding how OpenAI builds and releases new models.
Groq adapts after Nvidia deal reshaped AI cloud market
Groq is navigating major changes in the AI cloud market after a deal with Nvidia. The company has had to adjust its strategy as the competitive landscape shifts. Groq's situation highlights how partnerships and business moves can reshape the AI infrastructure industry.
China advances AI security as Hacking Cat and Vite server risks grow
China is pushing back on calls from US tech leaders to slow AI development. Foreign spokesperson Guo Jiakun warned that fearmongering could hurt global AI governance. Separately, pro-Ukraine group Hacking Cat has escalated cyberattacks on Russia using tools like Gorilla RAT. A mass-scanning campaign also targets exposed Vite development servers to steal cloud credentials. Security experts recommend updating software and rotating secrets to reduce risk.
Study examines trade-offs for vision-language models in document extraction
A new study looks at how vision-language models handle structured data extraction from business documents. Researchers tested eleven systems on a 750-document dataset and found that fine-tuning can boost open-source models above F1 0.98. The study also introduces a framework to help practitioners choose models based on quality, cost, speed, and governance needs.
AI will redefine leadership rather than replace it, experts say
AI is a leadership challenge, not just a technology issue, according to Forbes contributor Rodger deAnduncan. While AI can increase organizational output, only humans can decide what is truly worth producing. The article stresses that leaders must focus on human judgment, responsibility, and outcome quality as AI reshapes the workplace.
Job seekers quit AI interviews and blacklist companies over impersonal bots
About two-thirds of 1,200 job seekers walked away from a hiring process because of an AI interview, a survey found. Candidates said the bots felt cold, repetitive, and insulting. One 65-year-old IT director hung up early. An instructional designer with a eye movement condition said the AI told her to look straight ahead repeatedly. The ACLU has raised concerns about bias and discrimination in AI hiring tools.
Salesforce highlights AI strategy at Dreamforce and Investor Day
Salesforce kicked off its Dreamforce conference with CEO Marc Benioff's keynote and new AI product announcements. An investor day focused on long-term financial goals followed on Wednesday. The company's stock rose after its second-quarter earnings report. Signs showed Salesforce's AI strategy was gaining momentum.
Anthropic launches Salesforce in Claude plugin with 37 sales skills
Anthropic released the beta of Salesforce in Claude on September 15, 2026. The plugin brings seller accounts, opportunities, and pipeline data into Claude using existing Salesforce credentials. Claude can summarize accounts, update opportunities, log calls, and create follow-up tasks. GitLab, Siemens, and Legora have deployed the plugin, and 7,000 Salesforce sellers use it in their work.
Treasury's Scott Bessent says AI labs should not get liability exemptions
Treasury Secretary Scott Bessent spoke before the House Financial Services Committee about AI safety. He said the government should not give AI labs a liability exemption. His comments came days after Anthropic's Dario Amodei called for voluntary standards. Bessent also pushed for more open-source AI models built in the United States to compete with China.
Businesses adopt AI faster than they manage its risks, Nationwide survey finds
A Nationwide survey found that 60% of business owners say employees use public AI chatbots for work, but only 36% have written policies on AI use. About 31% of small and mid-market businesses say they were targeted by a generative AI scam in the past 12 months. Only 35% of owners have an up-to-date incident response plan. Eighty-two percent say they need more resources to protect against AI-enabled cyberattacks.
Sources
- Greg Brockman says OpenAI slowed cutting-edge AI work over safety
- Groq navigates AI cloud market shift after Nvidia deal
- China hits fast-forward on AI security, Hacking Cat shows its claws, Vite servers spill cloud secrets
- Beyond Accuracy: Robustness, Cost, and Governance Trade-offs for Vision-Language Models in Templated Document Extraction
- AI Won’t Replace Leadership—But It Will Redefine It
- Fed up with AI interviews, some job seekers are dropping out as candidates and blacklisting companies from consideration
- Salesforce Puts AI Strategy In Spotlight At Dreamforce, Investor Day
- Anthropic Releases Salesforce in Claude Plugin With 37 Sales Skills
- Treasury’s Scott Bessent says no liability exemptions for AI labs
- Nationwide Survey: Businesses Are Using AI Faster Than They Are Managing Its Risks
Comments
Please log in to post a comment.