Can organizations turn access to AI into reliable, useful capability?
Direction of power
Mixed-contested — proposed reconciliation
Window
Preserves; new routes Open
Confidence
Medium
Movement since August
More funded expertise and shared services; durable learning and exit remain unproven
End-of-September assessment
Implementation is acquiring more public and intermediary support. Mitacs offers subsidized expertise inside organizations. USAi moves beyond a free pilot to a continuing federal shared service. Queensland opens a funded manufacturing service route through ARM Hub. Professional accountability rules assign responsibilities to clinicians, lawyers and employers even when they use AI. These developments make organizational learning and integration more accessible.
They coexist with provider-controlled platforms, outside consultants and subsidy-dependent implementation. A placement target is not completed institutional expertise. Shared procurement is not proof that an organization can switch models without losing data, workflows or accumulated learning. Local training and jobs can be valuable even while a delivery center remains dependent on a small set of clients or vendors.
Baseline reconciliation requiring editorial approval: August's monthly scoreboard says Leaning dispersion. The current Lane 7 master says Mixed-contested and its September weekly text compares against Leaning concentration, omitting the published August baseline. This draft proposes Mixed-contested as the cumulative September judgment because the public implementation routes are concrete but lasting internal capability and exit remain incompletely demonstrated. That is an explicit analytical recalibration and repair of contradictory baselines, not evidence of a measured September deterioration. The September movement itself is toward more public and shared entry routes.
Other evidence. Türkiye's government-wide architecture entered into force August 18. DDD's center began operating August 1; its September announcement does not make the opening a September event. DDD's local-entry training pathway is real in the attributed account, while ownership, funding and transferable-skills outcomes remain unresolved. The public-capacity diagnostic below distinguishes experimentation and training from procurement memory, staff retention and replaceable systems.
What would change this assessment. Delivered placements and services with retained expertise; documented total cost and independent evaluation; common interfaces proven to support migration; internal teams retaining procurement and integration knowledge; or evidence that vendor-specific subsidies and consultants become the enduring operating layer.
Further implementation evidence. Britain's September 24 justice report describes live tools and internal delivery teams, including more than 1.5 million transcribed meetings. It is the ministry's retrospective self-report, not an independent outcome evaluation or proof that every deployment began in September. IMF Working Paper 26/147, published July 10 and recovered in September, models uneven geographic value from one provider's task-use data. Its labor-cost equivalent is potential time-saving value, not observed earnings, GDP or captured revenue. Australia's supplier-readiness programme was announced August 28; September deadlines do not establish selection or delivered bank contracts. See the unscored evidence register for dates, retained dispositions and source links.
Supporting scored events
The complete relevant event list follows. Primary and secondary references preserve one underlying event and score. Source holds and later commencement dates remain explicit; publication does not resolve them.
September scored actionSeptember 30, 2026 evidence cutoff
Qualcomm and ASUS launched a Pharmaceutical AI Agent field-deployment program for more than 50 demonstration pharmacies across southern Taiwan. The system uses the open-weight gpt-oss-20b model, public Taiwan FDA drug information, OCR and retrieval components, and local/on-premises inference on donated AI endpoint devices to support medication-safety and pharmacy workflows.
Limits and interpretation: Do not describe the deployment as compressing gpt-oss-120b into 20B. OpenAI released gpt-oss-120b and gpt-oss-20b as distinct models; the Taiwan program specializes and deploys the 20B variant. Field validation is beginning, demonstrated performance at scale remains limited, and the deployment still depends materially on Qualcomm hardware and ASUS/Taiwan implementation infrastructure.
Amazon announced a Selling Partner plugin that connects its seller tools to Amazon Quick and, in a U.S.-store beta, to Claude. The relevant users are businesses selling through Amazon. They can use a connected assistant to obtain information from Amazon’s seller tools and initiate supported actions, rather than having to conduct every interaction through Amazon’s own assistant interface. Amazon’s account says the connection includes human approvals, boundaries on access and audit trails. These controls matter because the assistant can help act on a seller’s business, not merely answer general questions.
The verified change is a live but limited external-assistant connection, with availability established through the provider’s announcement. This review did not independently test its performance, exportability or reliability. Amazon continues to operate the marketplace and the underlying seller services, and the connection does not establish a general entitlement for other assistants to join. Neither portable workflows, independent hosting, transferable accumulated context nor nondiscriminatory admission terms have been demonstrated.
OpenAI launched the Agents API in public beta, allowing developers to use OpenAI’s managed Codex agent harness while running agent work in OpenAI-hosted sandboxes, their own infrastructure, or integrated third-party environments including Cloudflare, DigitalOcean, Oracle and Vercel. The release supports MCP and multi-agent orchestration, and the underlying Codex harness is available under the Apache 2.0 license for independent inspection, copying and modification.
Limits and interpretation: The managed Agents API remains an OpenAI service optimized alongside OpenAI models, and the announcement does not establish model-provider neutrality for the managed service. It therefore improves execution portability and inspectability without ensuring full model substitutability or preventing OpenAI from becoming a dominant orchestration layer.
ServiceNow made prebuilt MCP servers generally available for IT Service Management, IT Operations Management, Configuration Management Database and Strategic Portfolio Management. Supported MCP-compatible clients including Claude, Microsoft Copilot and Gemini can invoke ServiceNow workflows while authentication, permissions, validation, workflow logic and execution controls remain inside ServiceNow.
Limits and interpretation: ServiceNow had already opened standards-compliant external agents to end-to-end workflow execution through its MCP Server Console and Action Fabric. The September 10 release makes that architecture materially easier to deploy through domain-specific, out-of-the-box servers, but it operationalizes an existing separation rather than creating it for the first time.
NSF awarded $35 million over five years to the San Diego Supercomputer Center, working with the Texas Advanced Computing Center, to establish the National Artificial Intelligence Research Resource Operations Center. The center will operate the national portal, integrate computing, data, models and tools from public and private providers, and provide support, training and coordination as NAIRR transitions from a pilot toward sustained infrastructure.
Limits and interpretation: A substantial portion of frontier capability beneath that public interface remains supplied by a concentrated commercial industry. NAIRR therefore increases contestability without eliminating dependence on concentrated upstream providers or proving full workload portability among them.
OpenAI committed $1 billion in subsidized Daybreak cybersecurity access, training, technical assistance and partnerships for state and local governments, utilities, health organizations, community banks, nonprofits and other essential-service defenders, including a U.S. pilot with MS-ISAC.
The Government of Canada launched a $13 million National AI Literacy Initiative with the Alberta Machine Intelligence Institute (Amii), providing free AI education for students, educators, workers and communities. The program targets up to one million post-secondary students and more than 50,000 K–12 educators and builds on an existing Amii consortium of 61 post-secondary institutions representing more than 900,000 students. Consortium expansion opened September 10, with course delivery and additional national modules scheduled to begin during September and the fall.
GSA and OpenAI executed a 27-month OneGov agreement, effective October 1, 2026 through December 31, 2028, that makes ChatGPT available to eligible federal, state, local and tribal governments with a 50% discount on consumption-based model use, no platform-access fee, no minimum order or spending commitment, multiple purchasing routes, government-data safeguards, training and deployment support. The agreement also includes most-favored-pricing protections, utilization and savings reporting, and a one-year checkpoint.
Limits and interpretation: The agreement was executed in September but takes effect October 1. September status is adopted procurement access, not completed implementation.
Canada’s federal government announced up to C$162 million over five years, beginning in 2026–27, for Mitacs AI Advantage. Evan Solomon, Minister of Artificial Intelligence and Digital Innovation, announced the funding in a release issued by Innovation, Science and Economic Development Canada. The money supports 10,000 co-funded work-placement opportunities for post-secondary students, recent graduates and postdoctoral fellows. Participants will work with Canadian businesses to help them adopt AI and develop innovations, while gaining paid practical experience themselves. Mitacs delivers the program and connects academic talent with organizations that need project expertise.
The practical mechanism is a subsidized placement working on an organization’s project, rather than simply access to an AI course or software subscription. The operational program accepts applications for adoption and research projects. Matching contributions, eligibility requirements and application review apply; the terms inspected during the September 26 review also include eligible municipal, nonprofit and Indigenous-government partners. The central federal announcement emphasizes Canadian businesses. The 10,000 placements are a program target, not a count of completed placements, and the funding ceiling is not evidence that all money has been disbursed.
GSA’s USAi platform completed its transition from a free pilot into a paid cost-recovery shared service for federal agencies. USAi is already used by 25 agencies, with 16 more expected to onboard by year-end, and provides a shared Chat interface, unified API and administrative/evaluation Console across models from multiple providers. Agency charges combine a platform fee with model usage passed through at cost.
Limits and interpretation: The service remains federal-only, agencies must now fund their own participation, and substantial authority over provider selection and platform architecture is concentrated in GSA. The new cost-recovery model could also produce uneven adoption across agencies with different budgets and implementation capacity.
Action/date record: September 2026; processed September 15. Approved score: Dispersion +2. Primary lane: 6. Secondary lanes: 7.
The foundation committed $1 billion to help lower-resource institutions and communities build and use AI capacity. The size, explicit equity purpose and implementation orientation make this more than a general pledge, while provider dependence and the durability of recipient capability remain follow-up questions.
Source hold: The approved ledger refers to an alert source archive, but an operative URL was not recovered from the records materialized for this review. The score is preserved; actor, precise action date, text and scope need source-archive reconciliation before this becomes fully verified public evidence.
September scored actionSeptember 30, 2026 evidence cutoff
Action/date record: September 2026; processed September 15. Approved score: Dispersion +1. Primary lane: 5. Secondary lanes: 7.
Salesforce and Google implemented Model Context Protocol connections across their agent and enterprise environments. The score recognizes a functioning standardized interface that lowers one layer of integration friction between major ecosystems; it does not establish full workflow, identity, memory or data portability.
Source hold: The approved ledger refers to an alert source archive, but an operative URL was not recovered from the records materialized for this review. The score is preserved; actor, precise action date, text and scope need source-archive reconciliation before this becomes fully verified public evidence.
September scored actionSeptember 30, 2026 evidence cutoff
Spain launched IA360 as a twelve-month government program combining AI oversight, public-interest deployment measures, business support, child protections and a formal process involving employers, unions and government in a proposed national AI social compact.
Verizon launched a national free-training initiative combining $50 million new funding with its existing $20 million Reskilling and Career Transition Fund. Multi-provider training is paired with community nonprofit support for workers, job seekers, educators and small businesses.
Limits and interpretation: The $70 million headline is not all new spending. Corporate sponsorship, curriculum independence, durable delivery, participation and transferable outcomes require follow-through. Training access is not itself worker decision authority.
Action/date record: September 22, 2026 announcement; processed September 23. Approved score: Dispersion +1. Primary lane: 5. Secondary lanes: 7.
Public preview provides managed persistent workspaces and governed tool access for competing agent harnesses and custom agent software. DigitalOcean identifies customers using the service.
Limits and interpretation: Preview is not general availability. Managed state, permissions and gateway services can create new dependencies; easy migration of accumulated work is not established. Performance and productivity figures are vendor-reported.
Action/date record: September 30, 2026. Approved score: Dispersion +2. Primary lane: 8. Secondary lanes: 1, 7. Stable record identity: D1001-22.
SB947, SB951 and AB1883 were signed on September 30. SB947 establishes human corroboration for discipline or dismissal principally based on an automated decision system, notice and information about a worker’s own data, and anti-retaliation provisions. Its principal employment protections commence July 1, 2027. Public authorities enforce the law; there is no private right of action, and collective-bargaining and federal-law exceptions qualify coverage. SB951 adds AI-replacement information to the final 60-day CalWARN notice, quarterly EDD summaries and a hiring report beginning January 1, 2028. AB1883 restricts specified neural and emotion surveillance with exceptions; it does not prohibit every form of monitoring. AB1331 bathroom privacy is wider digital-surveillance context, not a separate AI score.
Limits and interpretation: The coordinated package gives workers and public enforcers specified human-control, information and privacy mechanisms. The +2 recognizes the combined institutional consequence. It preserves the existing SB947 lifecycle; the earlier provisional +1 is superseded, not added. This is adoption rather than demonstrated compliance or improved labor-market outcomes.
Action/date record: September 30, 2026. Approved score: Dispersion +1. Primary lane: 1. Secondary lanes: 7. Stable record identity: D1001-23.
Official status confirms AB1979 chapter 854 and SB503 chapter 857, approved and chaptered September 30. Final enrolled text preserves licensed clinical judgment and prohibits autonomous performance of functions requiring a license, with supervised trainee and nonjudgmental documentation/communication qualifications. AB1979 brings covered health chatbots within medical-confidentiality law. SB503 assigns developers intended-use, training-data, evaluation and foreseeable-bias documentation and mitigation duties, while deployers must monitor and mitigate biased impacts. These are reasonable-effort/proportionate duties, not a promise of error-free medicine.
Limits and interpretation: Specified adopted obligations strengthen clinician authority and institutional accountability using existing licensing and confidentiality institutions. One coordinated +1 covers the two laws; no new regulator or individual general appeal right is inferred.
Action/date record: September 30, 2026. Approved score: Dispersion +1. Primary lane: 1. Secondary lanes: 7. Stable record identity: D1001-26.
SB574 chapter 858 was approved and chaptered September 30. The final enrolled act prohibits lawyers from delegating the practice of law to generative AI, requires personal verification of citations and protects nonpublic information under specified conditions. Arbitrators retain decision responsibility and disclose generated information drawn from outside the record. Broader non-AI arbitration amendments are not included in this AI score.
Limits and interpretation: Named professional responsibility and evidence-handling duties adapt existing institutions to AI use. They make consequential work accountable to humans and justify a modest +1 without assuming perfect enforcement or creating a new regulator.
Queensland commits A$750,000 to ARM Hub’s manufacturing AI uplift programme, with expressions of interest open until October 30. Flagship government contributions are A$150,000–250,000 each; 80% government/20% participant cash plus staff contribution, with at least one regional flagship. Support is delivered as engineering/services rather than cash grants, including data foundations, private Australian-cloud workspaces, integration and support. Eligibility includes 5–500 FTE, at least 50% Queensland-based employees and two years’ Queensland presence. The AI roadmap pathway does not require matched cash and is supported through the funded AI Adopt Centre.
Limits and interpretation: Committed finance and an available application pathway establish a concrete capability-building mechanism for smaller manufacturers. The +1 does not await selections, service commencement or measured skill retention. It is distinct from existing manufacturing microcredentials and is not a government public-service deployment.
Action/date record: September 2, 2026 reporting; formal plan entered into force August 18, 2026. Approved score: Dispersion +1. Primary lane: 7. Secondary lanes: 6, 1.
Türkiye’s 2026–2030 Artificial Intelligence Action Plan entered into force through Presidential Circular 2026/9 and establishes a government-wide implementation architecture including mandatory AI training with assessed micro-certificates, senior institutional AI administrators, three-year agency roadmaps, an Identify–Pilot–Scale cycle, at least 30 pilots across 10 ministries, pilot-to-scale pathways for critical public services, and certification/evaluation for high-impact systems.
Action/date record: Operational August 1, 2026; announced September 11; processed September 13. Approved score: Dispersion +1. Primary lane: 7. Secondary lanes: 8.
About 150 hires reportedly onboarded at DDD's first Indian Global Capability Centre. Foundational training lasts two to four weeks, followed by client-specific instruction. Expansion to 800–1,000 remains a target.
Limits and interpretation: DDD has a genuine U.S. 501(c)(3) entity alongside commercial operations and philanthropic support. Nonprofit ownership of every operating entity and freedom from vendor/government influence are not established. Indian ownership chain, client concentration and center-specific funding remain unresolved. Track transferable skills, wages, advancement, external placement and customer dependence, not simply headcount. New-center outcomes remain unproven.