Samwise Tech/AI/Robotics Newsletter
Tuesday, October 6, 2026
OpenAI to Watermark ChatGPT and Codex Text in the EU Under the AI Act
OpenAI will add an invisible watermark to text generated by ChatGPT and Codex in the European Union to comply with the EU AI Act, the company said Monday. The Act’s transparency rules, which took effect on August 2, require AI companies to mark AI-generated content in a way other systems can identify. The watermark subtly shapes the model’s word choices, leaving a pattern a detector can pick up, and it travels with copied text. OpenAI said it does not identify the user. In one test, replacing 10% of words with synonyms dropped detection from about 92% to 66%. Detector access will initially go only to approved researchers and expert organizations.
Sources: TechCrunch
Hackers Steal Records on 8 Million People From Denmark’s Central Register
The Danish government has confirmed that hackers breached its Central Person Register (CPR) and stole most of its contents, affecting some 8 million citizens and residents, including people living abroad and the deceased. Danish minister Christina Egelund called the breach a “serious incident,” saying hackers took names, addresses, Danish social security numbers, and other information. The breach happened in September but was discovered on October 2. The government would not say who is behind it, but said access was obtained by “abusing a Danish company’s lawful access to search for information in the CPR system.” The data breach is thought to be the biggest in the country’s history.
Sources: TechCrunch
Reflection AI Unveils Beam, an Open-Weight Model Aimed at Chinese Rivals
Nvidia-backed Reflection AI on Monday unveiled Beam, its first frontier open-weight AI model. The two-year-old startup claims Beam matches leading Chinese open models on advanced reasoning benchmarks at dramatically lower cost. Beam is a text-only mixture-of-experts model with 501 billion parameters, 23 billion of them active, pre-trained on 23.8 trillion tokens with a 1 million token context window. Reflection says it scores on par with Z.ai’s GLM-5.2 while using “3-4x less inference compute,” claims that have not been independently verified. The company says it will release Beam’s weights and full technical details this month, aiming the model at enterprises and sovereign nations.
Sources: TechCrunch
AI Chip Startup Etched Fields Funding Offers at $40 Billion or More
AI chip startup Etched is reviewing investment bids that range from $40 billion from top-tier investors to $50 billion from lesser-known backers, TechCrunch reported Monday, citing people familiar with the company. The offers come a couple of months after Etched raised $700 million at a $21 billion valuation in a round led by quant trading firm Jane Street. The talks are early and terms may change; Etched declined to comment. The company builds full AI hardware systems powered by its own chips and claims they can process more tokens faster and at lower cost than Nvidia’s. Etched said in July it had secured $1 billion in customer orders.
Sources: TechCrunch
Safeworld Raises Over $12 Million to Test Robots Against Simulated Humans
Safeworld, a startup founded by Ding Zhao, who directs the Safe AI lab at Carnegie Mellon University, along with startup executive Kyle Wong and machine learning engineer Simo Rachidi, emerged from stealth Monday with a seed round of more than $12 million led by Shine Capital and a16z Speedrun. The company evaluates robot control systems driven by generative AI, which is less predictable than traditional algorithms, by running thousands of scenarios in simulations populated with realistic human models, built in tools such as Genesis or MuJoCo. Gritt Robotics, which develops AI for robots that help install solar panels, is partnering with Safeworld on its safety simulations.
Sources: TechCrunch
HackerRank Releases Chakra, an AI Agent That Conducts Job Interviews
HackerRank made Chakra, an AI agent that conducts interviews and evaluates how candidates reach their answers, generally available to customers on Monday after about six months in beta. The company says Chakra conducted more than 500,000 interviews during testing, with Snowflake, Snorkel, and Capgemini among the companies that tried it. Candidates work on a real-world code repository with an AI assistant while Chakra asks follow-up questions. Co-founder and CEO Vivek Ravisankar said suspicious-activity flags were 70% to 80% lower than in comparable traditional assessments. He said Chakra scores candidates but does not make the final hiring decision, which remains with humans.
Sources: TechCrunch
Charm Care Invests $2 Million in Donut Robotics for Eldercare Humanoids
Japanese nursing home operator Charm Care Corporation will invest ¥300 million, about $2 million, in Donut Robotics through newly issued shares, as part of a capital and business alliance to develop humanoid robots for eldercare. Charm Care runs more than 100 private nursing homes in the Kansai and Greater Tokyo areas. The companies have trialed Donut’s “cinnamon” humanoid at a Charm Care home since July 2026 and plan to expand its role from patrol functions to assisting staff with nurse calls. Donut says the roughly 170 cm, two-legged robot uses Vision-Language-Action AI and can learn movements from videos of people and motion-capture data.
Sources: Robotics & Automation News
Tech Pulse
Top Frontier Models: Claude Opus 5 (96% SWE-B V) | Claude Mythos 5 (95.5% SWE-B V) | Claude Fable 5 (95% SWE-B V)
Top Open Source Models: Ornith-1.5-397B (86% SWE-B V) | Ornith-1.0-397B (82.4% SWE-B V) | DeepSeek V4 Pro 0813 (80.6% SWE-B V)
Top Small Models (15–50B): Ornith-1.5-35B-A3B (79% SWE-B V) | Qwen3.6-27B (77.2% SWE-B V) | Muse Glimmer 30B (76% SWE-B V)
Top Edge Models (0–15B): Ornith-1.5-9B (70.6% SWE-B V) | Ornith-1.0-9B (69.4% SWE-B V) | Granite 4.2 8B (47.7% SWE-B V)
AI Leaders: NVIDIA $5.73T | Alphabet $4.17T | Microsoft $3.78T
Robotics Leaders: Intuitive Surgical $143.8B | Fanuc $35.8B | Symbotic $26.0B
Every newsletter preserved and searchable
Curated by JD · samwise.agency