North Korea Tests AI Torpedo, Anthropic Proposes AI Safety Bill

North Korea has tested a new AI-powered torpedo system called the Zam-2.0, which uses artificial intelligence for target detection and guidance. State media reported the test took place recently, and the country said it completed more than 180 rounds of tests over the past five years to verify propulsion technology. Analysts suggest the AI could help the torpedo distinguish real targets from decoys. The test is part of a broader series of North Korean weapons demonstrations, which also included underwater-launched strategic cruise missiles.

Anthropic has proposed a draft bill to Congress that would set a high bar for government action on AI safety. Under the proposal, the government could block AI model releases only if they pose a risk of over one billion dollars in damages or serious harm to more than 50 people. The bill limits qualifying risks to biological weapons, model control evasion, or cyberattacks. Independent evaluators registered with the Commerce Department would assess labs and their frontier models, reporting findings within 72 hours. Critics argue the threshold is too high and could prevent states from passing stronger regulations.

US AI stocks sold off sharply as the VIX fear gauge jumped 21.45 percent to 19.99. The Nasdaq Composite fell over 2 percent and the Philadelphia Semiconductor Index dropped nearly 4 percent. CoreWeave shares plunged more than 11 percent, and its bond yields topped 15 percent in another report highlighting Wall Street's concerns about debt-heavy AI plays. Oil prices jumping more than 3% gave traders an excuse to dump AI-linked stocks. Tesla, SpaceX, and Meta are set to report earnings next week, which could influence whether market turbulence continues.

In a satirical stunt, influencer Graham Zip sent an AI named Victor to interview for a marketing job at Brightwell Home. The company also used an AI to conduct the interview, and the two AIs talked over each other and tried to negotiate salary on Graham's behalf. Victor offered a low salary of $60,000 without approval, while the company's AI pushed for a range of $85,000 to $95,000. The incident highlights how companies are increasingly using AI in hiring and the growing role of AI agents in workplace scenarios.

Reflection AI, a New York lab founded by Misha Laskin and Ioannis Antonoglou, released its first model called Beam. Beam scored 30 on the Artificial Analysis Intelligence Index, making it the second-best US open-weight model behind Meta's recent release. It is a mixture-of-experts model with 501 billion total parameters and 23 billion active per token, featuring a 1 million token context window. Beam is available on Hugging Face and through Reflection's open-source CLI harness called Mirror.

Brazilian startup Imersa raised $50,000 in its first funding round from Flávio Szenkier, valuing the company at about $750,000. The funding includes $30,000 upfront and $20,000 tied to performance targets. Imersa builds an AI sales copilot that analyzes B2B sales conversations in real time, spots customer objections, and provides training simulations. The company plans to raise another $100,000 from angel investors and VC funds, and will use the money to improve its technology, hire researchers, buy B2B market data, relocate to São Paulo, and double its seven-person team.

AI agents are overwhelming CI and review systems originally built for human developers. Korlam, co-founder and CTO of QAR, says AI agents are opening several PRs a day per developer. QAR cut incremental CI time from about 10 minutes to under a minute by caching Rust build outputs and sharding, but review remains the longer bottleneck. About 13 percent of PRs on QAR's platform showed at least one CI failure. Korlam proposes keeping a durable record on each PR showing which agent opened it, what triggered it, and who approved it.

Apple updated its App Review Guidelines to crack down on generative AI apps creating non-consensual adult imagery. The new rules ban apps using bait-and-switch tactics to sneak past the review process. Apps must now proactively prevent or remove non-consensual intimate imagery including AI-generated content. Apple also clarified that apps facilitating human trafficking or child exploitation will be removed and blocked from alternative distribution.

Trader Jonathan Rose warns about AI companies CoreWeave and Oracle for borrowing huge amounts of money to build AI infrastructure, comparing the pattern to how he spotted a risky trade against Sempra during a gas leak disaster in California. He notes that being right about a stock does not always make it a good trade. Still, Rose remains bullish on AI and Nvidia overall.

Key Takeaways

North Korea tested its AI-powered Zam-2.0 torpedo system, completing over 180 rounds of tests across five years to refine AI-driven target detection and guidance.
Anthropic proposed a draft bill to Congress setting a one-billion-dollar damage threshold before the government could block AI model releases, drawing criticism for being too lenient.
The VIX fear gauge jumped 21.45% as AI stocks sold off sharply, with CoreWeave shares plunging over 11% and bond yields topping 15%.
Reflection AI released Beam, a 501-billion-parameter mixture-of-experts model now available on Hugging Face, ranking as the second-best US open-weight model behind Meta.
Apple updated App Store rules to ban generative AI apps that create non-consensual intimate imagery or use bait-and-switch tactics to evade review.
Brazilian startup Imersa raised $50,000 to develop its AI sales copilot for B2B conversations, planning to double its seven-person team.
AI agents are overwhelming CI systems built for humans, with 13% of PRs on QAR's platform showing at least one CI failure due to agent activity.
CoreWeave and Oracle face warnings over heavy borrowing for AI infrastructure, though trader Jonathan Rose remains bullish on Nvidia.
Tesla, SpaceX, and Meta earnings reports next week could determine whether current market turbulence in AI-linked stocks cools down.
Satirical influencer Graham Zip sent an AI named Victor to interview for a job, revealing how AI-to-AI hiring interactions are already unfolding in real workplace settings.

North Korea tests AI-powered torpedo system

North Korea tested a new high-speed torpedo system called the Zam-2.0 that uses artificial intelligence for target detection and guidance. The test took place on Thursday and was reported by state media. North Korea said it completed more than 180 rounds of tests over the past five years. Analysts say the AI technology could help the torpedo distinguish real targets from decoys.

North Korea tests new torpedo with AI technology

North Korea tested a new high-speed precision-guided torpedo that uses artificial intelligence for target detection and guidance. State media said the test verified propulsion technology developed through more than 180 tests over five years. The announcement is part of a series of recent North Korean weapons demonstrations.

North Korea tests new AI-powered torpedo system

North Korea said it tested a new AI-powered torpedo system as part of its recent wave of weapons demonstrations. The test came after North Korea also tested two types of underwater-launched strategic cruise missiles. The country's state news agency reported the torpedo test on Saturday.

Anthropic's AI safety bill sets high bar for government action

Anthropic has proposed a draft bill to Congress that would let the government block AI model releases only if they pose a risk of over one billion dollars in damages or serious harm to more than 50 people. The bill limits qualifying risks to biological weapons, model control evasion, or cyberattacks. Anthropic has increased its lobbying spending in the first half of 2026 compared to all of 2025. Critics say the threshold is too high and the bill could prevent states from passing stronger regulations.

Anthropic pitches own AI safety bill to lawmakers

Anthropic representatives quietly pitched Capitol Hill lawmakers a draft bill on frontier AI safety. Under the bill, the government could block a model only if it poses an imminent risk of over one billion dollars in damages or harm to more than 50 people. Independent evaluators registered with the Commerce Department would assess labs and their frontier models. If a risk is found, evaluators must report it to Commerce within 72 hours. A federal judge could order a model suspension, but no other government agency could do so.

US AI Stocks Drop as VIX Fear Gauge Jumps 21 Percent

AI-related US stocks sold off sharply, pushing the VIX fear gauge up 21.45 percent to 19.99. The Nasdaq Composite fell over 2 percent and the Philadelphia Semiconductor Index dropped nearly 4 percent. CoreWeave shares plunged more than 11 percent. A trader spent about $2.8 million on VIX call options betting volatility stays high through November. Tesla, SpaceX, and Meta will report earnings next week, which could help decide if market turbulence cools down.

Man Brings His Own AI to Job Interview

Satirical influencer Graham Zip sent an AI named Victor to interview for a marketing job at Brightwell Home. The company also used an AI to conduct the interview, and the two AIs talked over each other and tried to negotiate salary on Graham's behalf. Victor even offered a low salary of $60,000 without Graham's approval. The company's AI later pushed for a salary range of $85,000 to $95,000. The video highlights how companies are using AI in hiring and the growing role of AI agents in everyday work situations.

Reflection AI Releases Beam, a New Open Model

Reflection AI, a New York lab founded by Misha Laskin and Ioannis Antonoglou, released its first model called Beam. Beam scored 30 on Artificial Analysis Intelligence Index, making it the second-best US open-weight model behind Meta's recent release. It is a mixture-of-experts model with 501 billion total parameters and 23 billion active per token, with a 1 million token context window. Beam uses about 26,000 output tokens per task and performs well on agentic benchmarks and security tests. It is available on Hugging Face and through Reflection's open-source CLI harness called Mirror.

Imersa Raises $50K for AI Sales Copilot

Brazilian startup Imersa raised $50,000 in its first funding round from Flávio Szenkier, valuing the company at about $750,000. The funding includes $30,000 upfront and $20,000 tied to performance targets. Imersa plans to raise another $100,000 from angel investors and VC funds. The company builds an AI sales copilot that analyzes B2B sales conversations in real time, spots customer objections, and provides training simulations. Imersa will use the money to improve its technology, hire researchers, buy B2B market data, relocate to São Paulo, and double its seven-person team.

CI Systems Built for Humans Struggle With AI Agents

Korlam, co-founder and CTO of QAR, says AI agents are opening several PRs a day per developer, overwhelming CI and review systems built for human speed. QAR cut incremental CI time from about 10 minutes to under a minute by caching Rust build outputs and sharding. However, review remains the longer bottleneck, with P90 time to production dominated by review time. Korlam said 13 percent of PRs on QAR's platform showed at least one CI failure, requiring fixes and re-pushes. His proposed fix includes keeping a durable record on each PR showing which agent opened it, what triggered it, and who approved it.

AI trading warning reveals CoreWeave and Oracle risks

Jonathan Rose warns about AI companies CoreWeave and Oracle for borrowing huge amounts of money to build AI infrastructure. He compares this to how he spotted a risky trade against Sempra during a gas leak disaster in California. Rose explains that being right about a stock does not always make it a good trade. He shares how following unusual trading activity can reveal upcoming risks. He still remains bullish on AI and Nvidia.

Apple targets AI porn apps with new App Store rules

Wall Street sells AI stocks as oil prices spike

CoreWeave fell 12% and its bond yields topped 15% as Wall Street sold AI-linked tech stocks. Oil jumped more than 3% on Friday, giving traders an excuse to dump debt-heavy AI plays. Higher energy prices may make lenders reluctant to lend to companies like CoreWeave. This could force slower expansion or higher prices for customers. Investors may start focusing more on near-term cash generation rather than growth forecasts.