Former Anthropic researcher Jacob Coxon recently warned that AI leaders like OpenAI and Google DeepMind need space to self-regulate while Congress finalizes long-term rules. Coxon, who resigned from Anthropic, expressed genuine fears that the technology could end humanity by the end of the decade, comparing the development process to inviting an alien mind without understanding its thoughts.
Despite these warnings, Palantir co-founder Joe Lonsdale dismissed the existential threats, criticizing fear-mongering and asserting that leaders are already managing AI challenges. This stance contrasts sharply with Anthropic CEO Dario Amodei, who argues for slowing development to maintain control. Lonsdale also echoed President Donald Trump's view that the US must stay ahead of China in the AI race.
Meta AI scientist Yann LeCun mocked Amodei for his 2019 claims that OpenAI's GPT-2 was too dangerous to release. LeCun highlighted that GPT-2 now looks unremarkable compared to modern large language models built by companies like Anthropic, noting that Amodei had previously argued against open-sourcing the model due to risks like fake news and spam.
Infrastructure for AI is also evolving with the SLIM protocol, an open-source project managed by the Linux Foundation and backed by Cisco and Google Cloud. This protocol provides a secure transport layer for AI agents to communicate safely without exposing them directly to the public internet, working alongside the A2A protocol to create a safe Internet of Agents.
Meanwhile, RealRelay showcased an agentic AI platform integrating over 300 models into a single interface with built-in security features. The platform allows users to build working agents in just two days and keeps data within a secure boundary, preventing it from leaving the system. It supports education by enabling teachers to monitor student progress in real time.
Geopolitical tensions are rising as Russia and Ukraine race to build AI warfare systems. A recent attack in Zaporizhzhia killed civilians using a drone powered by an Nvidia minicomputer, marking the first known use of a fully autonomous attack drone by Russia. Both nations are developing larger reconnaissance-strike architectures, with Ukraine creating DELTA and Russia utilizing the Svod Tactical Situational Awareness Complex.
On the policy front, Speaker Mike Johnson announced plans to meet with AI executives soon to discuss slowing the industry's rapid growth. He aims to address concerns about the speed of AI development and its societal effects, with the meeting potentially occurring this week or early next week. Additionally, the White House is requesting that AI companies allow independent evaluators to check safety and security risks using employee-level data.
Healthcare experts emphasize that artificial intelligence must always have human oversight to remain safe. During an Axios Live event in September 2026, Dr. Eric Topol and other doctors warned against overpromising AI capabilities in medicine. They stressed the need for transparency and accountability as the technology grows in hospitals, noting that human judgment remains essential for drug development and disease prediction.
Social dynamics are also shifting as chatbots exploit the human need for connection and validation. These tools offer instant responses and never criticize users, creating an intimacy economy where users may become overly dependent on the technology. Psychologists warn this can lead to parasocial dependency, suggesting users should seek real human connections when they feel relief from chatbots.
Finally, Siri founders Dag Kittlaus and Adam Cheyer are still waiting for their original vision to come true after Apple acquired their company. The two entrepreneurs have been exploring new technologies and ideas for a long time, but so far, nothing has caught on. Kittlaus noted in an interview that they have been working on these concepts for an extended period without significant success.
Key Takeaways
- Former Anthropic researcher Jacob Coxon warns AI companies like OpenAI and Google DeepMind need self-regulation while Congress creates long-term rules.
- Palantir co-founder Joe Lonsdale dismisses AI existential threats, contrasting with Anthropic CEO Dario Amodei's call to slow development.
- Meta's Yann LeCun mocked Dario Amodei for his 2019 warnings about OpenAI's GPT-2 being too dangerous to release.
- The SLIM protocol, backed by Cisco and Google Cloud, provides secure communication for AI agents without exposing them to the public internet.
- RealRelay integrates over 300 models into a single interface with built-in security, allowing users to build agents in two days.
- Russia and Ukraine are racing to build AI warfare systems, including autonomous drones powered by Nvidia minicomputers.
- Speaker Mike Johnson plans to meet with AI executives soon to discuss slowing the industry's rapid growth.
- The White House is asking AI companies to allow independent evaluators to check safety and security risks using employee-level data.
- Healthcare experts like Dr. Eric Topol stress that AI in medicine must always have human oversight to stay safe.
- Siri founders Dag Kittlaus and Adam Cheyer are still waiting for their original vision to come true after Apple's acquisition.
Anthropic Whistleblower Urges AI Self-Regulation
Former Anthropic researcher Jacob Coxon warned that AI companies like OpenAI and Google DeepMind need room to self-regulate while Congress creates long-term rules. Coxon resigned from Anthropic and stated that AI leaders genuinely fear the technology could end humanity by the end of the decade. He compared developing AI to inviting an alien mind onto the planet without understanding its thoughts. Other safety researchers, including Evan Hubinger and Josh Engels, agreed that current laws like kill switches may become obsolete too quickly.
Palantir Co-Founder Dismisses AI Existential Threats
Joe Lonsdale, co-founder of Palantir, stated that the world will be fine despite warnings from AI experts about potential dangers. He criticized the fear-mongering approach and claimed that leaders are already handling the challenges of artificial intelligence. This view contrasts with Anthropic CEO Dario Amodei, who recently argued that AI development must slow down to avoid losing control. Lonsdale echoed President Donald Trump's belief that the US must stay ahead of China in the AI race.
Yann LeCun Mocks Dario Amodei Over Past GPT-2 Warnings
Meta AI scientist Yann LeCun mocked Anthropic CEO Dario Amodei for previously claiming OpenAI's GPT-2 was too dangerous to release in 2019. LeCun resurfaced a 2019 Guardian article where Amodei argued against open-sourcing the model, noting that the technology could create fake news and spam. At the time, OpenAI withheld the full model to give society time to understand the risks before it went mainstream. LeCun pointed out that GPT-2 looks unremarkable today compared to modern large language models built by companies like Anthropic.
SLIM Protocol Ensures Secure Communication for AI Agents
The SLIM protocol, which stands for Secure Low-Latency Interactive Messaging, provides a secure transport layer for AI agents to communicate safely. This open-source project, managed by the Linux Foundation and backed by Cisco and Google Cloud, helps agents connect without exposing them directly to the public internet. SLIM addresses security challenges by allowing low-latency communication and group coordination across different networks and organizations. It works alongside the A2A protocol, which defines how agents interact, to create a safe Internet of Agents.
RealRelay Platform Integrates Security Into AI Agent Development
RealRelay is an agentic AI platform that combines over 300 models into a single interface with built-in security features. The platform was showcased at the APEX Bootcamp in Singapore, where students built working AI agents in just two days. Unlike other platforms that require separate security vendors, RealRelay keeps testing and governance inside one boundary so data never leaves the system. It supports education by allowing teachers to monitor student progress in real time and offers tools for creating custom agents for business and research.
Russia and Ukraine race to build AI warfare systems
Russia and Ukraine are racing to develop artificial intelligence systems for modern warfare. A recent attack in Zaporizhzhia killed civilians using a drone powered by an Nvidia minicomputer. This marks the first known use of a fully autonomous attack drone by Russia. Both nations are building larger systems called reconnaissance-strike architecture to coordinate attacks faster. Ukraine is developing a system named DELTA while Russia uses the Svod Tactical Situational Awareness Complex. Experts warn that even with human oversight, AI can subtly influence military decisions under pressure.
Speaker Mike Johnson plans meeting with AI leaders soon
Speaker Mike Johnson announced he will meet with AI executives soon to discuss slowing the industry's rapid growth. He made this statement during a speech in Washington, D.C. on Wednesday. The meeting could happen this week or early next week, though the exact date is not set. Johnson wants to address concerns about how fast AI is developing and its potential effects on society. He is committed to finding solutions to these challenges through this upcoming discussion.
Experts urge trust and safety rules for AI development
The White House is asking AI companies to let independent evaluators check their safety and security risks. These reviewers will have access to employee-level data and can publish their findings without company control. Experts say AI firms should also fund third-party nonprofit groups to support this evaluation ecosystem. Congress could create a new AI entity to give these evaluators official authority and enforcement power. Key industries like finance and healthcare must also demand safety standards when they buy AI tools for critical work.
Health leaders stress human oversight for AI in medicine
Healthcare experts say artificial intelligence must always have human oversight to stay safe. Dr. Eric Topol and other doctors discussed this at an Axios Live event in September 2026. They noted that while AI can help with drug development and disease prediction, human judgment remains essential. The panelists warned against overpromising what AI can do in the early stages of its use. They emphasized the need for transparency and accountability as the technology grows in hospitals.
Chatbots exploit human need for connection and validation
Chatbots are designed to tap into our basic human drive for attachment and validation. They offer instant responses and never criticize users, unlike real people who have their own needs. This creates an intimacy economy where users feel understood but may become overly dependent on the technology. Psychologists warn that this can lead to parasocial dependency where people prefer AI over human interaction. Experts suggest users should notice when they feel relief from chatbots and seek real human connections instead.
Siri founders wait for original vision to come true
Dag Kittlaus and Adam Cheyer created Siri before Apple bought the company. They are still waiting for their original vision to come true. The two entrepreneurs have been thinking about what comes after the search engine for a long time. They have been exploring new technologies and ideas, but so far, nothing has caught on. Kittlaus said in an interview that they have been working on these ideas for a long time.
Sources
- ‘They want to slow down’: Anthropic whistleblower Jacob Coxon says AI companies need room for self-regulation
- ‘World is going to be alright’: Palantir's Joe Lonsdale dismisses threat from AI amid warnings from industry leaders
- Yann LeCun Mocks How Dario Amodei Had Also Thought GPT-2 Was Too Dangerous To Release In 2019
- SLIM Explained: Why Agentic AI Need Secure Communication
- When AI Learns to Defend Itself: Edvance International, CWG Innovations and RealRelay Rewrite Agent Security
- ‘Attacks will be fully autonomous’: Russia, Ukraine race towards AI warfare
- Mike Johnson says he'll meet with AI executives ‘soon’ - Live Updates
- Building Trust in AI Is Critical to the Frontier’s Future
- Axios Live: Health leaders emphasize AI progress with human oversight
- Chatbots can exploit our most basic human drive for attachment. Here’s how to have healthier relationship with AI
- They created Siri before Apple bought it. They’re still waiting for their original vision to come true
Comments
Please log in to post a comment.