OpenAI launches GPT-6.2 Astra and GPT-6.2 Sol for different kinds of work. Astra targets demanding professional tasks such as presentations, spreadsheets, documents, coding, website development, gaming and 3D design. Sol is aimed at everyday use and improves on GPT-6.1 Sol in reasoning, efficiency and multi-step work. Both use fewer output tokens, and are available to Plus, Pro, Enterprise and Edu users through ChatGPT Work and Codex, as well as developers through the API. Google’s Gemini 4 Neon, released November 19, is designed to match or outperform Gemini 4 Carbon on coding, reasoning and agent tasks. Google says Neon is three times faster and 60% cheaper in initial tests; it offers a one-million-token context window and four thinking levels. It is rolling out across the Gemini app, AI Studio and Gemini for Business, with paid Pro and Ultra access. Pricing is $3 per million input tokens and $15 per million output tokens, with cached input at $0.15.
Google Cloud is also combining high-performance computing with AI agents to help researchers choose simulations and avoid spending time on non-essential parameters. The approach uses purpose-built hardware and open-source tools including Slurm and Kubernetes. The article notes AI’s production use in weather forecasting on a National Oceanic and Atmospheric Administration platform, while cautioning that AI output can be inaccurate. Elsewhere, demand for AI infrastructure is lifting exports in Asia: Japan’s exports surged, while Chinese, Japanese and South Korean firms supply chips and other equipment. Economists warn that the region could become increasingly reliant on those hardware sales.
AI oversight and reliability remain concerns. A survey by the National Association of State Chief Information Officers covering officials in 33 states and territories found that 97% have AI policies or guidance, but just 44% of respondents say their programs can independently set enterprise-wide standards. More than half lack dedicated budgets and 75% cite insufficient staffing. NASCIO recommends clear authority, steady funding, coordination on data and security, and employee training. In California, school districts may set their own AI rules because state guidance is not a mandate. Districts vary in readiness; the state recommends explaining acceptable use and not relying solely on AI detectors for discipline. Sonal Patel urged schools to involve educators, counselors and communities, keep human judgment central and revisit policies.
New Zealand’s Ministry of Education faced criticism for error-filled AI videos explaining its refreshed curriculum. Materials on the official Tāhūrangi site showed a dancer with three arms, culturally jarring te reo Māori slogans and a map placing Taranaki in Wellington; similar materials appeared in roadshow slides. The ministry said the videos were intended to help educators prepare for curriculum changes. Separately, Homeland Security Secretary Markwayne Mullin suggested AI may have caused a nationwide FAA ground stop and outages at JPMorgan, Microsoft and United Airlines, citing a conversation with an unnamed technology contact but no technical evidence. His claim conflicts with Transportation Secretary Sean Duffy, who blamed contractor L3Harris and said AI was not responsible. No official technical finding has established the cause, and Representative Ted Lieu criticized the speculation.
AI is also raising business and regulatory questions. Ross Douthat describes a growing divide between technology leaders who see AI as a way to overcome human limits and bankers focused on profits. The article says the two worlds moved closer in 2025 and the first half of 2026, with an Anthropic IPO expected to be the biggest in history; Douthat considers tech leaders too optimistic and bankers too cynical, without settling which view is right. For financial firms, AI-assisted marketing creates questions about authorship, supervision, recordkeeping and copyright. FINRA’s 2026 guidance highlights those duties and the regulator has sought comment on modernizing Rule 2210. Firms may need to document how content was created, while Morgan Stanley’s September DMCA complaints over X posts containing research screenshots illustrate copyright concerns.
In healthcare, Elsevier upgraded ClinicalKey AI so it can access a patient’s full electronic health record and assess treatment options in real time. San Juan Regional Medical Center is using the tool to provide clinicians with evidence-based guidance alongside patient-specific data, with the stated aims of improving care and easing workforce constraints.
Key Takeaways
- OpenAI’s GPT-6.2 Astra targets complex professional work, while GPT-6.2 Sol focuses on everyday tasks; both are available through ChatGPT Work, Codex and the API, with fewer output tokens.
- Google says Gemini 4 Neon is three times faster and 60% cheaper than Gemini 4 Carbon in initial tests, with a one-million-token context window; listed API prices start at $3 per million input tokens.
- Google Cloud is pairing AI agents with high-performance computing, Slurm and Kubernetes to help researchers select simulations; the article notes AI forecasting use but cautions that output can be inaccurate.
- AI hardware demand is contributing to stronger Asian exports, including Japan’s surge, while economists warn of greater regional reliance on hardware sales.
- In a survey covering 33 states and territories, 56% of state AI programs lacked dedicated budgets and 75% cited insufficient staffing, despite 97% having AI policies or guidance.
- California’s AI recommendations leave school districts free to set local rules; officials advise clear acceptable-use guidance, human judgment and not relying only on AI detectors for discipline.
- New Zealand’s Education Ministry drew criticism for curriculum videos with visible AI errors, including a three-armed dancer and a map placing Taranaki in Wellington.
- Markwayne Mullin speculated without technical evidence that AI may have caused several outages; Transportation Secretary Sean Duffy blamed L3Harris, and no official finding has established a cause.
- Financial firms face AI authorship, supervision, recordkeeping and copyright questions; FINRA’s 2026 guidance addresses these duties, and Morgan Stanley filed September DMCA complaints over research screenshots on X.
- Elsevier’s upgraded ClinicalKey AI can use full patient records to assess treatment options in real time; San Juan Regional Medical Center is using it to give clinicians patient-specific, evidence-based guidance.
OpenAI releases GPT-6.2 Astra and Sol for different kinds of work
OpenAI launched GPT-6.2 Astra and GPT-6.2 Sol, saying both models handle demanding tasks better while using fewer output tokens. Astra is designed for complex professional work, including creating presentations, spreadsheets and documents, as well as coding, website development, gaming and 3D design. Sol is aimed at everyday tasks and improves on GPT-6.1 Sol in reasoning, efficiency and handling multi-step work. Both models are available to Plus, Pro, Enterprise and Edu users through ChatGPT Work and Codex, and to developers through the OpenAI API.
Google launches Gemini 4 Neon, three times faster than Carbon
Google launched Gemini 4 Neon on November 19, 2026, describing it as comparable to or better than Gemini 4 Carbon on coding, reasoning and agent tasks. The company said Neon was three times faster and 60% cheaper in initial tests, and offers a 1 million token context window with four thinking levels. It is rolling out through the Gemini app, Google AI Studio and Gemini for Business, with access for paid Google AI Pro and Google AI Ultra users. Google set prices at $3 per million input tokens and $15 per million output tokens, with cached input tokens costing $0.15 per million.
US Homeland Security chief suggests AI caused outages without evidence
Homeland Security Secretary Markwayne Mullin said artificial intelligence may have caused last Friday’s nationwide FAA ground stop and outages at JPMorgan, Microsoft and United Airlines. He based the suggestion on a conversation with one unnamed technology contact and offered no technical finding to support it. His comments conflict with Transportation Secretary Sean Duffy, who blamed contractor L3Harris for “incompetence” and said AI was not responsible. Representative Ted Lieu criticized the speculation, and no official technical finding has established a cause for the outages.
Survey finds state governments lack resources to manage AI
A National Association of State Chief Information Officers survey of AI officials in 33 states and territories found that many state governments lack the resources and authority to oversee AI. While 97% of respondents said their states have AI policies or guidance, only 44% said they had independent authority to set enterprise-wide AI standards. The survey found that 56% of state AI programs lack dedicated budgets and 75% of respondents cited insufficient staffing as a major challenge. NASCIO recommends clear authority for AI leaders, steady funding, stronger coordination on data and security, and more employee training.
AI hardware demand pushes Asian exports to record highs
Japan’s exports surged as the country benefited from its role in the artificial intelligence supply chain. Chinese, Japanese and South Korean companies have become major manufacturers of chips and other equipment needed for the global AI infrastructure buildout. Economists worry that East Asian countries may become increasingly reliant on these rising hardware sales.
New Zealand education ministry faces criticism over flawed AI videos
New Zealand’s Ministry of Education faced criticism for using error-filled AI videos to explain its new curriculum to teachers and school leaders. The videos appeared on Tāhūrangi, the ministry’s official curriculum website, and included a dancer with three arms, te reo Māori slogans described as culturally jarring, and a map that placed Taranaki in Wellington. Similar AI materials appeared in slides at a ministry roadshow about introducing the curriculum. The ministry said the videos were made to explain the refreshed curriculum and help educators prepare for changes.
Google Cloud links AI agents with high performance computing
Google Cloud is developing an approach that combines high-performance computing with AI agents to help researchers reach discoveries faster. The strategy uses purpose-built hardware, open-source tools such as Slurm and Kubernetes, and collaborative AI agents. Researchers can use AI to select which simulation candidates to test, reducing time spent on non-essential parameters and helping create new prototypes. The article also describes production use of AI in weather forecasting on the National Oceanic and Atmospheric Administration’s platform, while noting that AI output can still be inaccurate.
Ross Douthat examines the clash between AI idealists and bankers
Ross Douthat examines the debate between technology leaders who see AI as a way to overcome human limits and bankers who focus on making money from it. The article describes how AI and finance seemed to move closer together in 2025 and the first half of 2026, with an Anthropic IPO expected to be the biggest in history. Douthat argues that tech leaders are too optimistic and bankers too cynical about AI. He says the truth lies between their views, though the article does not give a clear answer about which side is right.
AI content creates authorship and copyright risks for financial firms
Financial firms face growing questions about who authored content when advisors use AI to write or edit marketing materials. AI detectors may treat fully generated writing, grammar edits, added research, and heavily revised AI drafts alike, so firms may need to document how each piece was made. FINRA’s 2026 guidance points to duties involving supervision, communications, recordkeeping, and fair dealing, and the regulator has sought comment on modernizing Rule 2210. AI can also make it harder to trace third-party research and other material used in content, raising copyright concerns. The article cites Morgan Stanley’s September DMCA complaints over posts on X that included screenshots from its research.
Elsevier upgrades ClinicalKey AI with full patient record access
Elsevier announced an upgrade to ClinicalKey AI at HLTH 2026 in Las Vegas. The clinical decision support tool can now access a patient’s full electronic health record and assess treatment options in real time. San Juan Regional Medical Center is using it to give clinicians quicker access to evidence-based guidance alongside patient-specific health record data. The stated goals are to improve care and ease workforce constraints.
California AI Guidance Leaves Room for Local School Policies
At the California IT in Education conference, Sonal Patel discussed how school districts are using the state’s recommendations on AI. California’s guidance is not a mandate, and districts can create their own policies on academic integrity, student privacy, and age-appropriate AI lessons. Some districts are still unsure which tools teachers may use, while others have committees, staff training, and updated guidance. The state recommends clearly explaining acceptable AI use and not relying only on AI detection software for discipline. Patel urged districts to involve educators, counselors, and community members, keep human judgment central, and revisit their guidance over time.
Sources
- OpenAI launches GPT-6.2 Astra and Sol artificial intelligence models
- Google launches Gemini 4 Neon with Carbon-level performance at 3x speed: Features, pricing and availability
- Homeland Security Chief Blames AI for Outages on 'One Guy's' Word, Contradicting DOT
- NASCIO Survey Finds Gaps in AI Governance, Data Readiness Across State Governments
- AI hardware demand lifts Asian exports to record highs
- Ministry of Education criticised for ‘insulting’ AI-generated videos explaining new curriculum
- HPC meets AI agents as Google Cloud targets time to discovery
- Ross Douthat: Are the Bankers or the Tech Bros Right About AI?
- AI Authorship Raises New Risk for Financial Firms
- HLTH 2026: Elsevier upgrades ClinicalKey AI with full patient EHR context
- CITE26: California's AI Guidance Meets Local Realities
Comments
Please log in to post a comment.