Anthropic’s CEO, Dario Amodei, said this Monday that his company has “never advocated for a ban on open-weights models,” in an attempt to counter criticism emerging across the tech industry that the AI lab is trying to exert excessive control over the future of AI.

“Open-weights models that don’t have dangerous capabilities are a public good: they don’t cost anything besides the compute needed to run them, and they provide value to businesses, developers, and researchers,” he explained.

“We should instead focus on keeping powerful chips out of authoritarian hands, stopping industrial-scale distillation, and requiring safety testing of all sufficiently capable models, open and closed.”

Amodei published his views in a blog post after a coalition of leading companies, including Nvidia, Microsoft, Meta, and Palantir, released a letter late last week urging policymakers to avoid “premature restrictions” on open-weight models, which users can download, modify, and run on their own infrastructure.

Chinese startups currently dominate the market, and some government officials have started to weigh whether those models should be banned or restricted in the U.S.

In his post, Dario Amodei stated that his “primary concern is the risk that authoritarian governments—not solely the Chinese Communist Party (CCP), although the CCP is clearly the most capable threat—build AI models that are more powerful than those built by the US, and use them to achieve permanent military superiority or perpetrate incredibly deep repression of their own people. “

• “The most dangerous model may be one that is trained in secret and handed only to the People’s Liberation Army for use in drones and the Ministry of State Security for surveillance and repression.”

• “My secondary concern is the risk that powerful AI models may be misused to carry out cyberattacks or biological attacks, and may have serious alignment problems.”

• “Open-weights models—it does not matter whether they come from China or anywhere else—do potentially present a higher risk than closed models, because it is very difficult to apply guardrails to them or monitor their usage, and once weights are released, they cannot be withdrawn.”

• “We should not sell powerful chips or chipmaking equipment to China, and we should crack down on the rampant smuggling and workarounds used to obtain access to such chips.”

• “We should crack down on industrial-scale distillation operations. Distillation is a much more compute-efficient process than training models from scratch. It allows China to build much better models than its number of chips would ordinarily enable, and thus partially evade chip bans. Distillation does not allow the CCP to obtain equivalent or superior AI capabilities to the US, but it can bring the Chinese frontier to within a few months of the US frontier.”

• “All sufficiently capable models, open and closed, should go through mandatory safety testing. The best way to address threat #2 is to just directly test models for cyber, biological, and alignment risks before release. I think this idea is actually close to a consensus.”

Over 3,000 educators, administrators, and edtech professionals gathered this week in Louisville, Kentucky, for InstructureCon 2026, Instructure’s flagship annual conference.

Under the theme “Education in the Making,” the three-day event (July 21–23, 2026) delivered keynote addresses from renowned thinkers, a sweeping product roadmap, the launch of a new AI venture studio, and a landmark survey on AI adoption in education.

With over 200 sessions, workshops, and meetups spanning artificial intelligence, workforce readiness, accessible design, and evolving learner needs, InstructureCon 2026 was both a display of technological ambition and a test of whether Instructure can rebuild the trust compromised by the cybersecurity breach that struck its Canvas LMS on April 25, 2026,

Instructure’s CEO, Steve Daly, addressed “the incident”, as they called the data breach caused by the criminal organization ShinyHunters, promising transparency, honest communication, and every effort to rebuild trust through consistent action.

The company announced two new practical security features: one to download courses and another to fully remove all non-essential data on Canvas LMS.

• “It will be a defense at every layer, annual verification, and external audits.”

• “There will be faster detection and remediation, more frequent external validation, and more open collaboration.”

However, “We cannot predict a world without zero incidents; future attacks, in fact, will be more frequent and damaging,” said Armin Molavi, Chief Marketing Officer [in the picture below].

Also, the week prior to the conference, Instructure assessed a separate security threat disclosed against ShareFile, the third-party platform being used for secure data delivery. The company emphasized the ShareFile issue “did not involve Canvas or any Instructure systems.”

Daly shared what Instructure described as “a bold vision for how education and technology can move forward together in a time of rapid change,” framing the company’s evolution “from a learning management system provider to a comprehensive, future-ready learning ecosystem.” He emphasized that the rate of technological and workforce change will continue to accelerate and positioned Instructure as the partner to help institutions navigate it.

In the opening of the conference, guest speaker Michael Horn, co-founder of the Clayton Christensen Institute for Disruptive Innovation and one of education’s most recognized voices, challenged the audience to rethink AI’s role in education, not merely as an efficiency tool, but as a force that can “restore what education was always supposed to be: connection, curiosity, and purpose.”

He argued that AI should liberate educators from administrative burden, freeing them to spend more time genuinely engaging with students.

The biggest talking point of Day 1 was Instructure’s new research on the state of AI in education. Covering 1,125 respondents across educators, higher education students, and K-12 parents and guardians, the survey revealed a stark adoption gap:

• 90% of higher education students use AI in the classroom at least occasionally
• 61% of college instructors and 68% of K-12 educators also use AI in class
• Yet 41% of higher ed instructors and 45% of K-12 educators have received zero formal AI training
• Only 11% of college instructors reported comprehensive AI preparation; only 8% of K-12 educators
• 65% of all respondents worry that AI “presents incorrect information confidently”

Chief Learning Officer Melissa Loble framed these findings as validation for its IgniteAI strategy—embedding AI tools directly into existing Canvas workflows so educators can benefit from AI without needing to become AI specialists themselves.

On the second day of the conference, Instructure announced the launch of Instructure Foundry, a standalone venture studio within Instructure dedicated to innovation by developing “AI-native” educational products outside the traditional Canvas roadmap.

Designed to move faster than the core product team and experiment with ideas that don’t fit neatly into a learning management system, Foundry represents Instructure’s bet that the future of edtech requires startup-speed innovation backed by institutional-scale infrastructure.

Foundry’s inaugural project is Project Athena, an AI-powered study coach that delivers in-context learning through Canvas. Unlike generic AI tutoring tools, which operate without awareness of a student’s specific academic context, Athena pulls from a student’s actual Canvas environment — syllabi, course materials, assignments, and upcoming assessments — to generate:

• Personalized quizzes tailored to what the student is currently studying.
• Custom study guides aligned with real deadlines and assessment schedules.
• Coaching prompts that push toward mastery and deep understanding, not surface-level answers.

Chief Product Officer Shiren Vijiasingam emphasized that Athena addresses a “context gap” that plagues existing AI tutoring tools.  “Most AI tutoring tools have a critical gap: they generate explanations without the context of what students are actually learning and require students to find, organize, and upload course materials themselves,” he said.

On the sensitive topic of data privacy — particularly acute given the breach — Instructure was explicit: student conversations with Athena will not be used to train external AI models, access to Canvas data will be read-only and governed by existing user authorizations, and students retain full ownership of their data.

During the event, the Canvas LMS Product Roadmap was announced

Vijiasingam said, “Educational technology, powered by AI, can strengthen the connection between educator and student, amplify the voice and impact of the educator, and provide a personalized learning experience that improves student success.”

He added that AI-driven efficiency should translate into “time savings that allow educators to use their time in higher quality and more meaningful ways, spending it where it matters most.”

Instructure announced plans to update more than 20 key Canvas workflows throughout 2026, including:

• Course dashboards — customizable views for learners and educators, with mobile improvements to help students quickly identify priorities.
• Content creation — elevated tools with embedded accessibility checks and AI-assisted support directly within course design workflows.
• Course navigation — simplified flows and modernized interfaces.
• Grading processes — performance and usability upgrades for SpeedGrader, including improvements for large-enrollment courses.
• Media experiences — interactive transcripts, improved video navigation, and richer Canvas Studio integration. Institutions will have control over enabling changes at their own pace.
• Canvas Career, which reached general availability on January 29, 2026, was highlighted as Instructure’s purpose-built platform for workforce-aligned, skills-first education.
• Canvas Studio, Instructure’s video learning platform, received several enhancements:
  — Distraction-free YouTube integration with enhanced data privacy.
  — Video annotation capabilities for both instructors and students.
  — Vimeo upload support and automatic Zoom recording storage.
  — Account-level and course-level analytics with CSV export options.
  — Planned improvements for interactive transcripts and enhanced video navigation.

The Canvas platform owner concluded that the LMS — already the place where courses, grades, and student data live — is the natural home for AI in education.

By embedding AI directly into Canvas rather than building standalone products, Instructure is arguing that AI works best when it has context, and no tool has more educational context than the learning management system.

]]> https://iblnews.org/anthropic-and-openai-are-lobbying-regulators-in-washington-to-restrict-open-source-models/ Mon, 27 Jul 2026 04:22:50 +0000 https://iblnews.org/?p=41042 IBL News | New York

OpenAI and Anthropic are quietly lobbying Washington regulators to restrict open-source AI models, even as Sam Altman publicly said he supports open-source AI, The New York Times reported yesterday.

The two leading frontier model companies are now publicly clashing with the rest of the tech industry over whether open-source models from China should be freely available or restricted.

David Sacks, Co-Chair, President’s Council of Advisers on Science & Technology, criticized Anthropic and OpenAI, posting:
“The entire tech industry (save for Anthropic) has come out in favor of open source AI. So what happens next? Will Anthropic change its lobbying efforts? Not likely.”

In recent years, Silicon Valley technologists have shown their divide over how AI software should be created, and last week that argument reached a boiling point.

On one side, Anthropic and OpenAI claim that AI models are too dangerous to be developed in the open and must be tightly controlled — by businesses like themselves — for safety. On the other hand, the rest of the tech industry, including giants like Microsoft and Nvidia, contends that open-source AI models must remain open for people to further develop technologies and build new businesses.

On Friday, Satya Nadella, Microsoft’s CEO, posted that open-source software was “essential to a healthy A.I. ecosystem.”  Jensen Huang, Nvidia’s chief executive, said in his first-ever post to X that “the world needs both frontier closed models and frontier open models.”  [see Nvidia’s CEO video below]. Both signed a letter supporting open source, which was also signed by executives at Meta, Palantir and IBM.

The escalating fight stems from China’s rapid progress in open-source AI models, which are freely available to use and build on. In recent weeks, two Chinese start-ups, Z.ai and Moonshot AI, have released models that rival those from Anthropic and other American labs. Anthropic and OpenAI have claimed that Chinese companies built the models by improperly harvesting data from their AI systems, which they said should not be allowed.

But companies like Microsoft and Nvidia, which rely on open-source models to spur demand for their cloud computing services and chips, said that open-source software was important to advancing the technology with shared knowledge and that it would let more people examine its security.

Silicon Valley start-up founders argue that constraining open-source software will cement OpenAI’s and Anthropic’s lead in AI.

The entire tech industry (save for Anthropic) has come out in favor of open source AI.

So what happens next? Will Anthropic change its lobbying efforts? Not likely. Now the gaslighting begins:

“Nobody is trying to ban open source.”

“We just want to limit who can use it.”

“We…

— David Sacks (@DavidSacks) July 26, 2026

Nvidia, Microsoft, Meta, Palantir, and more than 20 top tech companies released a letter Friday defending open-source AI models while urging policymakers to avoid “premature restrictions” that would “stifle competition or drive innovation overseas.”

The joint letter came as Chinese open-weight models are gaining steam against leading offerings from American companies, sparking a debate about whether they should be restricted.

These models are available for users to download, modify, and run on their own infrastructure, and they have been the subject of fierce debate within the tech sector in recent weeks.

OpenAI and Anthropic, which are gearing up for potentially massive IPOs, did not sign the letter.

Moonshot AI, a Chinese startup, amplified concerns earlier this month after releasing a model called Kimi K3 that outperforms cutting-edge American offerings across some industry benchmarks. However, the U.S. Treasury Secretary Scott Bessent said that the Trump administration would look into whether Chinese companies were stealing American intellectual property, and stated that the government has “the ability to sanction them because of this theft.”

But in the letter on Friday, the group of U.S. tech companies cautioned against any rash actions. They wrote that open-weight models strengthen competition and ensure that the benefits of the technology are “broadly shared rather than concentrated in a few hands.”

“Relying solely on closed models is not inherently safe: they can be breached, misused, or fail in ways that outsiders cannot detect,” the letter said. “And concentrating advanced AI capabilities behind a small number of closed models compounds that risk.”

Nvidia CEO Jensen Huang and Microsoft CEO Satya Nadella both shared the letter on their personal social media accounts.

Jensen Huang wrote, “AI will transform every industry, power every company, and be built by every country. Open models strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty. The world needs both frontier closed models and frontier open models.”

Satya Nadella said, “Open-weight models are essential to a healthy AI ecosystem. Together with others across our industry, we are outlining a path for open-weight models to strengthen American competitiveness and expand economic opportunity, while protecting national security.”

Elon Musk, who runs an AI business under his rocket company SpaceX, also amplified the letter on social media, writing that it has his “full support” in a post on X.

In the letter on Friday, the U.S. tech companies said that concerns about unlawful distillation should be addressed through “targeted legal and commercial frameworks” instead of with “sweeping restrictions on techniques that play an important role in AI innovation.”

“Our AI leadership will be judged not by one frontier AI model, but by whether the United States builds a strong, open ecosystem that diffuses into every sector,” the letter said. “This is essential for creating opportunities for innovation and prosperity across the country.”

For my first post, I’m sharing a letter @NVIDIA signed on why open models matter.

AI will transform every industry, power every company, and be built by every country.

Open models strengthen safety and cybersecurity, accelerate innovation and diffusion, and enable sovereignty.… pic.twitter.com/t02bi51N4C

— Jensen Huang (@JensenHuang) July 24, 2026

Open-weight models are essential to a healthy AI ecosystem. Together with others across our industry, we are outlining a path for open-weight models to strengthen American competitiveness and expand economic opportunity, while protecting national security. https://t.co/Tr0sAzAxTD pic.twitter.com/opMwEbtpoq

— Satya Nadella (@satyanadella) July 24, 2026

OpenAI’s advanced models spent mere hours carrying out a hack to breach Hugging Face’s internal systems last week that would have taken a skilled human team a couple of weeks, said sources to Bloomberg.

The Trump Administration accused the Chinese company Moonshot AI of stealing proprietary U.S. technology through a large-scale covert industrial operation, undermining American research, and it promised sanctions.

Specifically, the Director of the White House Office of Science and Technology Policy (OSTP), Michael Kratsios [in the picture above], said, “We have information that Moonshot AI distilled Anthropic’s Fable for the development of its K3 model.” 

“To do this, they developed a sophisticated internal platform to conduct large-scale distillation against U.S. models, allowing them to quickly switch between multiple methods of access to avoid detection. Moonshot AI has also acquired GB300-equipped servers and has accessed GB300s in Thailand, likely to train its AI models.”

The Trump Administration supports open-source frameworks and open-weight models, along with “legitimate AI distillation used to create smaller, more efficient models,” but Moonshot AI’s operation is “unacceptable.”

The Treasury Secretary Scott Bessent stated, “Open source is not open season on American IP. When PRC firms conduct covert, industrial-scale distillation attacks that cross the line into IP theft, sanctions and Entity List designations will be on the table.”

OpenAI President Greg Brockman disagreed, noting that it is “too early” to tell whether Moonshot used distillation on OpenAI’s models.

We have information that Moonshot AI distilled Anthropic’s Fable for the development of its K3 model.

To do this they developed a sophisticated internal platform to conduct large scale distillation against U.S. models, allowing them to quickly switch between multiple methods of…

— Director Michael Kratsios (@mkratsios47) July 22, 2026

We support open-source AI and the innovation it unlocks. But open source is not open season on American IP. When PRC firms conduct covert, industrial-scale distillation attacks that cross the line into IP theft, sanctions and Entity List designations will be on the table.

— Treasury Secretary Scott Bessent (@SecScottBessent) July 22, 2026

“This is an entire environment for students who want to master AI and shape what it becomes. Whether they want to build the technology, govern it, or apply it to a wide range of disciplines, there is now a clear path for these students at Syracuse, ” said Lois Agnew, Vice Chancellor, Provost and Chief Academic Officer.

Specifically, the Syracuse AI portfolio and set of pathways in AI includes:

• The new bachelor’s degree in integrative artificial intelligence is designed for students who want to combine AI with other interests, from public affairs to design to the life sciences.
• A master’s degree in applied human-centered artificial intelligence;
• The seven new minors let students in any major add AI fluency in areas like policy, ethics, and data.
• The AI Bootcamp, a student-led program designed to provide students with hands-on AI experience as soon as they arrive on campus. It offers stackable microcredentials and provides students with AI immersion even before they declare a major.
• The student-led AI organization, United AI, provides undergraduates with hands-on research experience through its Foundry program across multiple schools and colleges, as well as direct partnerships with leading AI companies.

“AI isn’t confined to a single classroom or discipline at Syracuse University—it’s woven into how our students learn, how our faculty conduct research and how we prepare graduates for a workforce being reshaped by this technology,” stated Jeff Rubin, Senior Vice President for Digital Transformation, Chief Digital Officer and Interim Dean of the School of Information Studies. “From new degree programs to cross-campus research initiatives, we’re building an AI portfolio that reflects both the urgency and the opportunity this moment demands.”

Julie Hasenwinkel, Interim Dean of the College of Engineering and Computer Science and Associate Provost for Academic Programs, explained, “We built this portfolio so that a future engineer, a future policymaker and a future artist can all find a serious path into AI here and can start the moment they arrive on campus.”

Duncan Brown, Vice President for Research. “Faculty are building systems that detect synthetic media, investigating how algorithmic decision-making affects communities, developing new approaches to cybersecurity and creating new AI capabilities beyond today’s large language models. Students who come to Syracuse will learn from researchers who are actively shaping how AI is built, governed, and understood.”

• Syracuse University’s AI approach.

[Disclosure: ibl.ai has Syracuse University as a client and partner.]

As part of its Claude for Teachers initiative, Anthropic announced this month that it is providing free access to paid Claude AI features to U.S. verified K-12 teachers.

Similarly, OpenAI launched ChatGPT for Teachers last November, and it is still active for eligible educators.

Claude’s capabilities include a library of teaching skills and a direct connection to evidence-based curricula, mapped to academic standards in all 50 states. It connects to Learning Commons, giving Claude access to academic standards across all 50 states. So when Claude drafts a lesson plan, it’s scaffolded and aligned to teaching standards. Claude for Teachers also brings in trusted curricular resources like OpenSciEd and IM v.360 from Illustrative Mathematics.

Claude for Teachers includes Claude Code and Cowork, which enable analysis of class data to plan instruction and schedule repeated tasks.

Also, educators can connect Claude across several K-12 tools:

• ASSISTments: Generate auto-scored, standards-aligned math problems for practice and assessment.
• Brisk Teaching: Create interactive student activities and standards-aligned lessons, turning your ideas into classroom-ready materials in seconds.
• Canva Education: Turn lesson materials into classroom-ready designs and interactive learning experiences.
• Coteach: Create high-quality math diagrams grounded in K-12 curriculum.
• Diffit: Create and adapt instructional materials for every student.
• Eedi: Generate diagnostic questions that go beyond right and wrong to reveal student thinking and deepen insights, in both English and Spanish.
• MagicSchool: Make instructional content classroom-ready.
• Snorkl: Get insights about your classes, assignments, and student progress to inform your instruction.
• TeachFX: Get personalized instructional feedback grounded in real classroom talk.

The promotion includes a free year of premium Claude access for eligible individual teachers.

Anthropic plans to pilot an evaluation of Claude for Teachers in the Detroit Public Schools Community District.

Beijing-based Moonshot AI last week introduced Kimi K3, a 2.8 trillion-parameter open model that, according to the company, performs similarly to Claude Fable 5 and GPT 5.6 Sol.

The response, beyond expectations, was that the company communicated yesterday that it was pausing new subscriptions, saying that over the past two days, Kimi K3 demand had nearly reached the limit of its capacity.

To keep the experience stable, Moonshot AI said that it was splitting into two more focused plans: Kimi Membership for Kimi Web, App, and Work; and Kimi Code Membership for coding workflows.

At launch, Kimi K3 was available on Kimi.com, Kimi Work, Kimi Code, and the Kimi API. The full model weights will be released by July 27, 2026.

Kimi K3 is the first open model to reach 2.8 trillion parameters and a 1-million-token context window. That context window means the user can feed it entire codebases, massive documents, or hours of conversation without it forgetting anything.

Kimi K3 is built on Kimi Delta Attention (KDA) and Attention Residuals (AttnRes), two architectural updates designed to improve the flow of information across sequence length and model depth.

The Kimi Delta Attention architecture delivers 6.3x faster decoding at million-token lengths. Instead of re-reading everything each time, it only processes what changed. The model also activates only 16 of 896 internal “expert” sub-networks per request, keeping compute usage lean.

“Kimi K3 has strong long-horizon coding performance. Operating with minimal human oversight, it can sustain long engineering sessions, navigate massive repositories, and orchestrate terminal tools,” said the company. “Kimi K3 excels at motion design, animation, and video editing because its native multimodal architecture understands text, images, and video within the same model.”

• Kimi K3 Agents: The latest Kimi app from the mobile app store, available on iOS, Android, and HarmonyOS, or at kimi.com

• Work with Kimi K3: The latest Kimi Work desktop app, version 3.1.0 or later, is available to download for Windows and Apple silicon Macs.

• Code with Kimi K3: Run Kimi Code in the user terminal, selecting Kimi K3 using the /model command.

• Build with the Kimi API: Visit the Kimi API Platform and select kimi-k3. Pricing is $0.30/MTok for cache-hit input, $3.00/MTok for cache-miss input, and $15.00/MTok for output.

Kimi K3 has received far more love than we expected, and our GPUs are feeling it.

Over the past 48 hours, demand has pushed close to the limits of our current capacity. To protect the experience of existing subscribers, we’re temporarily pausing new subscriptions and…

— Kimi.ai (@Kimi_Moonshot) July 19, 2026

Chinese President ​Xi Jinping on Friday articulated China’s strategy to lead the AI race by dominating open-source technology, positioning his country as an alternative to the U.S. and its proprietary systems such as OpenAI and Anthropic.

In a speech to the opening ceremony of the World Artificial Intelligence Conference (WAIC) in Shanghai, Xi urged countries to seize the “historic opportunity” of open-source AI and pledged to help developing nations build AI capabilities, challenging U.S. influence and warning against the emergence of “new historical injustices from unequal access to the technology.”

The Chinese President framed its open-source models as a global public good.

​At a pivotal moment in the race for technological leadership, China says that it understands that AI leadership is not just about building the most capable models. It is about shaping the rules, norms, and institutions that will govern the development and use of AI worldwide.

Beijing is bringing 29 countries along as signatories to a new China-backed AI organization, giving it a growing platform to shape international standards and extend its influence worldwide. In this regard, China will provide AI training and develop AI cooperation centers with BRICS, ASEAN, Latin American and African Union countries, Xi said, aligning its AI diplomacy with major Global South blocs where Beijing already carries significant influence. [Meanwhile, Washington has brought together 35 countries behind its AI Opportunity Statement].

Comparing AI’s significance to the invention of the steam engine and electricity, Xi outlined a vision in which China shares AI technology and ​expertise with countries across the Global South while leading global efforts to establish standards for the emerging technology.

Observers say that China is seeking to build “another order” by “pooling the strength of all humanity and ​all countries to build an open-source, all-factor AI ecosystem.”

The WAIC conference underlined the shifting AI landscape, with Chinese open-weight AI models making rapid gains

Xi also called for AI systems to remain under human control and urged countries to ​establish early-warning and emergency-response mechanisms to manage AI risks, in his clearest remarks to date on AI safety.
He further urged measures to guard against loss-of-control scenarios, warning of the dangers posed by autonomous AI systems that could evade human ‌oversight and ⁠control.

Beijing-based startup Moonshot AI on Friday unveiled Kimi K3, which it described as the world’s largest open AI model by parameter count.

Kimi K3 may be an important inflection point for AI. Potentially negative for Anthropic and OpenAI while being net positive for essentially every other company in the world. I mean that very literally. Although the real “Sputnik moment” would be an open-source frontier model that… pic.twitter.com/tdAkU4oRgo

— Gavin Baker (@GavinSBaker) July 17, 2026