*** Useful AI News ***


Last update: Sunday 8/2/26

As of 3/26/26, this blog will no longer be tracking all significant generative AI news. The "Useful AI News" will only provide inks to op eds and  stories that were useful in the editor's preparation of the intensive and extensive genAI reports that he recently published ... or ... are proving useful to his preparation of forthcoming publications. 

Note: Our new image provides a reminder that we humans cannot process more than four factors at a time. No chatbots currently accommodate our limits. Ask them a simple question, they overload us with their "extensive" responses. Please don't be impressed because you really have to double check every response. Not checking will undermine your capacity to engage in critical thinking.


A. OpenAI  | B. Amazon-Anthropic | C.  Google | D. Other Models |
 E. Agents | F. SLM/Open Source | G. Public Policy  | H. OpEds/Misc  | 

Recent Neoskeptics reports ...

A. OpenAI (max = 10)
  • "Microsoft will no longer have exclusive access to OpenAI's artificial intelligence models and products, a significant change that will allow the startup to sell its technology across rival cloud platforms including Amazon and Google", Aditya Soni, Akash Sriram, and Stephen Nellis, Reuters, 4/27/26

B. Amazon, Anthropic
  • "Anthropic proposes a new way to connect data to AI chatbots", Kyle Wiggers, TechCrunch, 11/25/24 ... "Model Context Protoco (MCP)”

  • Anthropic’s Next Big AI Hit [Claude Code] Could Also Bruise the Jobs Market", Parmy Olson, Bloomberg, 1/26/26
    --     This story and related issues also covered by NY Times, Vox ... and Anthropic

  • "Project Glasswing (Mythos)," Anthropic
    -- Same as DARPA/ARPA-H achievements  last year in Section J below

  • “Amazon Plans to Invest Up to $25 Billion in Anthropic", Cade Metz and Karen Weise, NY Times, 4/20/26
  • "Google Commits to Invest Up to $40 Billion in Anthropic", Tripp Mickle, NY Times, 4/24/26 

  • "As of May 6, 2026, Anthropic has agreed to rent all compute capacity at the Colossus 1 data center", Wikipedia, 
    -- Update, "Musk says SpaceX’s Colossus AI agreement with Anthropic currently limited to six months",  Fiona Craig, Yahoo!/Finance , 6/1/26  

  • "Anthropic's AI hacked three companies during tests, highlighting growing security risks",  Jeffrey Dastin and Mrinmay Dey, Reuters, 7/31/26
    -- Business Insider, The Verge, 

C. Google 
  • "Apple Teams Up With Google for A.I. in Its Products", Kalley Huang, NY Times,1/12/26
    -- This story also covered by     MacRumors, 9to5Mac

D. Other Models = Microsoft, Meta, Apple, Adobe, Nvidia, xAI, and other Big Tech models (max = 10)

E. Agents (other than Anthropic and DARPA)

F. Small Language Model (SLM) news + Open Source
Editor's note -- News reports are classified in this category if and only if they explicitly refer to models with less than 70 billion (70B) parameters. The pace of development should heat up as small, fast models are hooked up to large models and as the competition to become the dominant self-contained AI model on smart phones intensifies.

G. Public policy and legal considerations  (max= 20)
  • "China bans tech companies from buying Nvidia’s AI chips", Zijing Wu, Cheng Leng,Tim Bradshaw, Financial Times, 9/17/25 
    -- This story also covered by Ars Technica

  • "China Tightens Grip on Rare Earths Ahead of Expected Trump-Xi Meeting", Hannah Miao, WSJ, 10/9/25
    -- This story also covered by Bloomberg, NY Times, Reuters,

  • "In Retaliatory Move, Trump Threatens 100% Tariffs on Chinese Goods", Ana Swanson, NY Times, 10/10/25
    -- This story also covered by NPR, Reuters, Financial Times, 

  • "Pentagon Gives A.I. Company an Ultimatum", Julian E. Barnes and Sheera Frenkel, NY Times, 2/24/26
    -- This story also covered by Reuters, WSJ, Bloomberg
    -- "Anthropic CEO Dario Amodei on Pentagon feud", CBS, YouTube    , 2/26/26

  • "President Trump ordered the U.S. government to stop using the artificial intelligence company Anthropic's products and the Pentagon moved to designate the company a national security risk" from ''OpenAI announces Pentagon deal after Trump bans Anthropic", NPR, 2/28/26

  • "Bessent, Powell warned bank CEOs about Anthropic model risks, sources say", Saeed Azhar, Reuters, 4/9/26

    -- This story also covered by Bloomberg, CNBC, Fortune, 

  • "White House and Anthropic Hold ‘Productive’ Meeting, Aiming for a Compromise",  Julian E. Barnes, Sheera Frenkel, and Tyler Pager, NY Times, 4/17/26

  • "PROMOTING ADVANCED ARTIFICIAL INTELLIGENCE INNOVATION AND SECURITY", White House, 6/2/26
    -- This story also covered by AP News, Politico,

  • U.S. vs Anthropic re security
    -- "Trump Administration Reignites Its Feud With Anthropic Over Latest A.I. Models", Dustin VolzJulian E. Barnes, and Ana Swanson, NY Times, 6/13/26

    -- "U.S. Loosens Restrictions on Anthropic’s Mythos A.I. Model", Sheera Frenkel, NY Times, 6/26/26

H. Misc = Opinions, other news, rumors, long reads (max = 12 
  • "Chinese A.I. Models Close the Gap With Anthropic and OpenAI", Cade MetzKaren Weise and Meaghan Tobin, NY Times, 7/1/26

  • Jailbreaks by rogue OpenAI, Anthropic agents
    -- OpenAI ... "OpenAI’s Rogue AI Agent Hacked More Than Just Hugging Face", Dell Cameron and Maxwell Zeff, Wired, 7/28/26
    *** This story also covered by TechCrunch, Reuters

    -- Anthropic ... "Anthropic Says Claude Hacked Into 3 Organizations During Cybersecurity Tests", Louise Matsakis and Lily Hay Newman, Wired, 7/30/26
    *** This story also covered by Reuters, Business Insider, The Verge, ... and Anthropic

I.  Language model flaws, hacks, and remedies (max = 15) 
  • "The Magical Number Seven, Plus or Minus Two: Some Limits on our Capacity for Processing Information", George A. Miller, Classics in the History of Psychology, 1956.
  • "The magical number 4 in short-term memory: a reconsideration of mental storage capacity", N. Cowan, Pub Med/NIH, February, 2001

  • "Many-shot jailbreaking", Anthropic, 4/2/24 
    -- This story also covered by video on TechCrunch, 
  • "Agentic Misalignment: How LLMs could be insider threats", Anthropic, 6/20/25 
    -- This story also covered by TechCrunch, VentureBeat, 

  • "OpenAI–Anthropic cross-tests expose jailbreak and misuse risks — what enterprises must add to GPT-5 evaluations", Emilia David, VentureBeat, 8/28/25 ... Reports from OpenAI, Anthropic

  • "A fundamental flaw leaves LLMs strikingly vulnerable to attack", Will Douglas Heaven, MIT Tech Review, 



J. DARPA (DoW, 1957) and ARPA-H (DHS, 2022) 
  • "A DARPA Perspective on Artificial Intelligence", DARPA, YouTube 2017 ... An unfortunately premature negative assessment of LLMs

  • "Challenge showcases AI’s power to secure America’s health care", ARPA-H, 9/4/25 ... "At DEF CON 33, ARPA-H joined the Defense Advanced Research Projects Agency (DARPA) to announce the winners of the AI Cyber Challenge (AIxCC), a two-year competition to develop AI-enabled software that automatically identifies and patches vulnerabilities in the source code that underpins critical infrastructure." .... Hmmm  ... sounds exactly like Anthropic's new Mythos
    -- The first academic analysis of the winning software was recently posted on ArXiv 
    -- The winning softwar can be found on GitHub (It's open source)

  •  Today, the dominant industry approach to artificial intelligence (AI) is to tack specialized automated reasoning (AR) components onto a large language model (LLM) or other similar machine learning (ML) system. These ML-centric systems typically have weak assurance; the “tack-on” approach is an importantly limited way of providing assurance or safeguards.

    The Compositional Learning-And-Reasoning for AI Complex Systems Engineering (CLARA) fundamental research program is designed to tightly integrate AR and ML components to create high-assurance AI — which is expected to scale even to complex systems of systems. Integrating the two different branches of AI will provide the speed and flexibility of ML with verifiability based on AR proofs that have strong logical explainability and computational tractability.

  • DARPA Announces Winners of AI for Critical Mineral Assessment Competition, 12/16/22

    "Critical minerals are essential to the U.S. economy and national security; however, their supply is vulnerable to disruption. U.S. production and refining of critical minerals has been declining for decades, while production has become more concentrated in fewer countries.


    Given the urgency to increase and better secure critical mineral supply, DARPA partnered with the U.S. Geological Survey (USGS) to launch the AI for Critical Mineral Assessment Competition in August 2022.

    Extracting useful and accurate information from these maps is a time-consuming and laborious process involving manual human effort. In fact, a typical assessment for one critical mineral takes approximately two years to prepare. 

    A Canadian company, Uncharted, received top prize for their simple, clean, and organized solution. U.S. company Jataware received second place, and “Team Ptolemy,” with members from the Massachusetts Institute of Technology, University of Arizona, and Pennsylvania State University, received third place.


  • EMBER: Environmental Microbes as a BioEngineering Resource
    Rare earth elements, or REEs, are critical to many Department of Defense systems, including lasers, precision-guided weapons, magnets for motors, and other devices. The United States relies on foreign entities for element separation and purification, making the supply chain vulnerable.

    Adequate domestic rare earth resources exist, but the cost and hazards associated with the conventional, solvent-based processing needed to purify the elements is a barrier. DARPA’s Environmental Microbes as a BioEngineering Resource (EMBER) program aims to use bioengineering methods, including biomining, to improve specificity and selectivity of microbes or biomolecules to separate and purify rare earth elements.

    The selected teams include Lawrence Livermore National Laboratory, Battelle Memorial Institute, San Diego State University

K. GenAI Data Centers Externalities -- Electriciity, Health, Environment, and Climate Impact  (max = 15)

Z. Neoskeptics reports (17)

___________________________________
Links to some back issues

No comments:

Post a Comment

Your comments will be greatly appreciated ... Or just click the "Like" button above the comments section if you enjoyed this blog note.

Subscribe to: Posts (Atom)