Rep. Greg Casar, a Democrat from Texas, is pushing for the U.S. to take the lead in global AI diplomacy. He argues the country should focus on ending the AI arms race rather than winning it. He plans to speak at the Pro-Human Assembly in Washington alongside Bernie Sanders, advocating for international agreements on dangerous AI as a core goal of foreign policy. Separately, Casar urged AI whistleblowers and activists to speak more forcefully about the threats from advanced technology, saying the time for careful language is over. He spoke at the conference run by the Future of Life Institute and noted he has heard from experts who fear describing how dangerous the technology has become because they worry about sounding crazy.
The House Science Committee announced it will take a 30-day pause to assess the risks of artificial intelligence. The committee chair says the fast pace of AI development raises serious concerns, and the pause will allow time to work with experts to find ways to reduce risks. Meanwhile, Van Jones, a CNN contributor and former Obama administration official, discussed potential AI regulation on CNN, saying the world is on the edge of a major technological revolution and explaining why regulation matters for the future.
In personnel news, Josh Engels left Google DeepMind's AGI safety team after just three weeks. He joined AI evaluation group METR because he believes advanced AI could cause immense harm within five years. Engels said he enjoyed his work and had even turned down offers from Anthropic and OpenAI, but decided the stakes were too high to stay. On the product front, Perplexity launched Portable Computer for Windows, a local AI agent for multistep tasks that requires Nvidia RTX GPUs with at least 24GB of VRAM, supporting cards like the RTX 3090, 4090, and 5090.
Cisco announced new Splunk innovations designed to help customers scale AI safely and affordably, addressing data quality, governance, and security challenges without costly data migration. Zoom introduced an AI-powered revenue operating system with tools called Engage and Forecast, which create multichannel sales sequences and convert deal signals into real-time forecasts. Dutch chip startup Axelera AI signed multiple contracts to supply chips to AI factories and launched its second chip on September 15. ThreatCanary launched an Australian AI-native security platform combining External Attack Surface Management, API Security, and Autonomous Offensive Security, naming Marco Delgado as Co-Founder and Chairman.
A separate analysis calls for government agencies to adopt AI black boxes, similar to aircraft flight recorders, to track decisions and ensure accountability when AI acts in mission-critical operations. An agentic control plane would enforce rules on which agents can run and access systems, the article says, noting that clear policies and audit trails are needed before giving AI systems more autonomy.
Key Takeaways
- <ul><li>Rep. Greg Casar calls for U.S. to lead global AI diplomacy and end the AI arms race rather than win it</li><li>House Science Committee plans a 30-day AI stand-down to assess risks and work with experts</li><li>Josh Engels left Google DeepMind's AGI safety team after three weeks, joining METR over concerns advanced AI could cause harm within five years</li><li>Engels turned down offers from Anthropic and OpenAI but decided the stakes were too high to stay</li><li>Perplexity launched Portable Computer for Windows, a local AI agent requiring Nvidia RTX GPUs with at least 24GB of VRAM</li><li>Cisco introduced new Splunk tools to help customers scale AI safely without costly data migration</li><li>Zoom launched an AI-powered revenue operating system with Engage and Forecast tools for sales teams</li><li>Axelera AI signed multiple chip supply contracts and launched its second chip on September 15</li><li>ThreatCanary launched an Australian AI-native security platform and named Marco Delgado as Co-Founder and Chairman</li><li>Government agencies are urged to adopt AI black boxes and agentic control planes for accountability in mission-critical operations</li></ul>
Casar calls for US to lead global AI diplomacy
Rep. Greg Casar, a Democrat from Texas, plans to call for the U.S. to lead in artificial intelligence diplomacy. He says the country should work to end the AI arms race rather than try to win it. He will speak at the Pro-Human Assembly in Washington alongside Bernie Sanders and others. He believes international agreements on dangerous AI should be a key goal of American foreign policy.
Casar urges AI whistleblowers to speak out on dangers
Rep. Greg Casar called on activists and whistleblowers to speak more forcefully about threats from advanced AI. He said the time for careful language is over. He spoke at the Pro-Human Assembly, a conference run by the Future of Life Institute. Casar said he has heard from experts who are afraid to describe how dangerous the technology has become because they worry about sounding crazy.
House Science Committee plans 30-day AI stand-down
The House Science Committee will take a 30-day pause to assess the risks of artificial intelligence. The committee chair says the fast pace of AI development raises serious concerns. The pause will allow time to work with experts and find ways to reduce risks. The chair says this careful approach is important for the country and future generations.
Google DeepMind staffer quits over AI safety fears
Josh Engels left Google DeepMind's AGI safety team after three weeks. He joined AI evaluation group METR because he believes advanced AI could cause immense harm within five years. He said he enjoyed his work and had even turned down offers from Anthropic and OpenAI. He still decided to leave because the stakes were too high.
Van Jones discusses potential AI regulation
Van Jones, a CNN contributor and former Obama administration official, explains the potential regulation of artificial intelligence. He says the world is on the edge of a major technological revolution. He focuses on how AI might be regulated to keep it safe. The discussion aired on CNN and covers why regulation matters for the future.
Perplexity launches local AI agent for Windows PCs with RTX GPUs
Perplexity released Portable Computer for Windows, a local AI agent for multistep tasks. It requires Nvidia RTX GPUs with at least 24GB of VRAM. Local work does not consume credits, but cloud tasks need permission. Supported cards include RTX 3090, 4090, and 5090.
Cisco introduces new Splunk tools to scale trusted AI
Cisco announced new Splunk innovations that help customers scale AI safely and affordably. The tools address data quality, governance, and security challenges. Customers can scale AI applications without costly data migration. The advancements support AI wherever data already lives.
Government agencies need AI black boxes for accountability
Government agencies need clear policies and audit trails before giving AI systems more autonomy. The article calls for an AI black box like an aircraft flight recorder to track decisions. An agentic control plane would enforce rules on which agents can run and access systems. This ensures accountability when AI acts in mission-critical operations.
Zoom launches AI-powered revenue operating system
Zoom introduced an AI-powered revenue operating system with new tools called Engage and Forecast. Engage creates multichannel sales sequences and prioritizes follow-up. Forecast converts deal signals into real-time forecasts. Zoom also launched ZRA Elite and Common Room, with three pricing tiers available.
Axelera AI wins supply deals and launches second chip
Dutch chip startup Axelera AI signed multiple contracts to supply chips to AI factories. The company also launched its second chip on September 15. The deals and new chip mark a key expansion for the European startup in the AI hardware market.
ThreatCanary launches Australian AI security platform
ThreatCanary launched an Australian AI-native security platform that combines External Attack Surface Management, API Security, and Autonomous Offensive Security. The Sydney company named Marco Delgado as Co-Founder and Chairman. Founders Matt Flannery and Andrew Horton say the platform helps organisations see how weaknesses can be linked into attack paths. It includes modules for vulnerability intelligence, dark web monitoring, and asset management. The company highlights API security as a key focus because custom interfaces often lack regular updates.
Sources
- Casar to call for US to lead AI diplomacy
- Casar urges AI whistleblowers to ‘speak the truth’ as worst-case predictions mount
- How to keep AI from going rogue? Start with a 30-day stand-down.
- Google DeepMind Staffer Quits, Says AI Could Cause 'Immense Harm' In 5 Years
- Van Jones explains potential AI regulation
- Perplexity’s local AI agent comes to Windows, but only for RTX GPUs with at least 24GB of VRAM — Portable Computer brings AI for multistep tasks to compatible PCs
- Cisco Delivers Trusted AI at Scale Through New Splunk Advancements
- Why government agencies need a ‘black box’ for AI systems
- Zoom unveils AI-powered revenue OS with new sales tools
- European chip startup Axelera wins AI factory supply deals and launches second chip
- ThreatCanary launches sovereign AI security platform
Comments
Please log in to post a comment.