A Guth Labs publication ↗
Guth News

Policy

Policy

AWS previews enterprise prompt guardrails in Quick

Amazon Web Services announced a preview of enterprise guardrails for AWS Quick on November 20. The controls let administrators choose what users can put in prompts, including whether to block selected types of personal information, phrases or topi…

Sources: aws.amazon.com
  1. Agents

    Eclipse announces Enclave 1.0 sandbox for autonomous coding agents

    The Eclipse Foundation announced Enclave 1.0 on November 19, an open-source runtime that puts AI coding agents in isolated, policy-controlled containers. The foundation’s announcement says the MIT-licensed software supports Claude Code, Codex, Ope…

    Sources: newsroom.eclipse.org, enclave.eclipse.dev
  2. Models

    GitHub adds Supabase, Box and six more Copilot connectors

    GitHub added eight service connectors to Copilot in public preview on November 18, including Supabase, Box and Calendly. The company’s release note says the integrations use GitHub-verified Model Context Protocol servers. The other additions are B…

    Sources: github.blog
  3. Guth News

    House committee requests NVIDIA and Commerce export-control briefings

    House Energy and Commerce Committee Chairman Brett Guthrie requested briefings from NVIDIA and the Commerce Department on November 18 about semiconductor-export controls and security safeguards, setting a December 2 deadline. The committee’s annou…

    Sources: energycommerce.house.gov, investor.nvidia.com
  4. Agents

    MIND expands data protection for Microsoft Copilot and Microsoft 365

    MIND announced generally available data-protection integrations for Microsoft Copilot and Microsoft 365 on November 18, extending its tools for finding sensitive information and enforcing restrictions across supported Microsoft applications, accor…

    Sources: prnewswire.com
  5. Policy

    Qodo previews policy-based code approvals and coding-session reviews

    Qodo introduced Auto Approval Workflows in beta and Agent Session Review as a research preview on November 18, adding policy-based approvals and a broader view of coding agents’ work, according to its announcement. Teams can set approval criteria …

    Sources: qodo.ai
  6. Agents

    Thrive announces managed AI governance service for December 3

    Thrive announced Managed AI Governance and Security services on November 18, with availability scheduled for December 3 rather than immediately. The offering combines governance technology, advisory work and the company's round-the-clock security …

    Sources: globenewswire.com
  7. Agents

    FlowForma announces Agent Trust hub for monitoring AI agents

    FlowForma announced Agent Trust on November 18, describing an available governance hub for monitoring AI agents across different software platforms. The Dublin company lists prebuilt connections for Microsoft Copilot Studio, Power Automate and n8n…

    Sources: prnewswire.com, flowforma.com
  8. Models

    Microsoft launches AI Transparency Initiative with more than 40 organizations

    Microsoft announced an AI Transparency Initiative on November 17, bringing together more than 40 organizations to develop a shared approach to verifying and governing AI systems. Participants named in the company announcement include Accenture, Ad…

    Sources: news.microsoft.com
  9. Guth News

    GitHub previews adaptive code reviews and secret detection

    GitHub added an Adaptive effort option to Copilot code review on November 17, letting the reviewer choose Lite, Balanced or Max based on a pull request's complexity and risk. The public preview shows which level was selected and respects the maxim…

    Sources: github.blog
  10. Models

    Perplexity launches Orchestrator API with shared billing and model failover

    Perplexity launched its Orchestrator API on November 17, combining access to proprietary and open-weight models with consolidated billing and provider failover. Its announcement lists support for OpenAI, Anthropic, Google and xAI models alongside …

    Sources: perplexity.ai
  11. Agents

    Cogent releases DefenseOS for AI-driven vulnerability management

    Cogent released DefenseOS on November 17, expanding its cybersecurity platform with tools that give AI agents organizational context and automate recurring vulnerability-management work. DefenseOS and Agent Automations are available to all Cogent …

    Sources: prnewswire.com
  12. Agents

    MLPerf adds agentic post-training to its AI hardware scorecard

    MLCommons released MLPerf Training v6.1 on November 16, adding a large-model reinforcement-learning workload to the benchmark suite as AI infrastructure expands beyond conventional pre-training. The new Qwen3.5-397B test uses Group Relative Policy…

    Sources: mlcommons.org, blog.google +1
  13. Agents

    Soteria pairs verified plans with runtime checks to keep AI agents on task

    A November 13 research paper proposes a different approach to agent guardrails: give the system a checked plan it can return to when an action is blocked, rather than merely telling it to stop. Soteria, by Mike He, Ankush Desai, Sharad Malik and A…

    Sources: arxiv.org
  14. Tools

    Philippines creates national AI council and adopts governance framework

    Philippine President Ferdinand Marcos Jr. signed Executive Order 131 on November 12, adopting an AI governance framework and creating a National Artificial Intelligence Council, according to the government’s Philippine News Agency. The signing was…

    Sources: pna.gov.ph, philstartech.com
  15. Policy

    Meta and TikTok challenge Indonesia’s proposed copyright rules alongside Google

    Meta and TikTok joined Google at an Indonesian parliamentary hearing on November 11 to challenge proposed copyright-law revisions covering news payments and artificial intelligence. The dispute concerns a bill under discussion, not an enacted AI c…

    Sources: aol.com, suara.com +1
  16. Policy

    French panel calls for pre-launch checks on consumer AI

    A report submitted to President Emmanuel Macron proposes stronger oversight of chatbots, including scrutiny of addictive design and hidden commercial influence. Guth News | Policy | Story date: November 10, 2026 An expert commission convened by Fr…

    Sources: elysee.fr, conseil-ia-numerique.fr +1
  17. Models

    Inkling 2 Small moves from hosted access to downloadable weights

    Thinking Machines Lab has released downloadable weights for Inkling 2 Small, changing the access story from its November5 hosted introduction. A clearly dated November12 update on the original announcement says the weights are available on Hugging…

    Sources: thinkingmachines.ai, huggingface.co +2
  18. Robotics

    Robbyant releases LingBot-Touch to help robot hands share tactile understanding

    Robbyant, Ant Group’s embodied-AI company, announced LingBot-Touch on November 10, offering a tactile perception layer intended to work across different robot hands and control models. The release targets an integration problem: touch sensors base…

    Sources: via.tt.se, arxiv.org +2
  19. Agents

    CapLite puts small-agent write permissions in code, with an explicit utility cost

    A November 12 preprint proposes a defense for small open-weight agents that does not ask their language model to author the security rules. CapLite derives policy from tool schemas, fixes write operations before reading tool results and checks con…

    Sources: arxiv.org
  20. Agents

    OpenAI opens Business Hub and customer-facing agents inside ChatGPT

    OpenAI introduced Businesses in ChatGPT in beta on November 10, giving companies a way to claim profiles, update their information and inspect aggregate engagement data. Eligible businesses can use Business Hub wherever ChatGPT is available; Busin…

    Sources: openai.com
  21. Agents

    Google's agent governance reaches outside its cloud, with a security tradeoff to understand

    Google has opened a limited preview of Agent Gateway for cross-cloud, extending centralized agent governance to CNCF-conformant Kubernetes clusters outside Google Cloud. Its November 11 release notes name Amazon EKS, Azure AKS, OpenShift and bare-…

    Sources: docs.cloud.google.com
  22. Policy

    Cloudflare makes AI security policy deployable with Terraform, with a migration catch

    Cloudflare announced Terraform management for AI Security for Apps on November 12, bringing zone-level activation and custom prompt topics into infrastructure-as-code workflows. Its announcement specifies Cloudflare provider 5.29.0 and two resourc…

    Sources: developers.cloudflare.com
  23. Agents

    ElevenLabs moves from better transcripts to agents that remember and follow through

    ElevenLabs is extending its speech business into the work around a conversation. At its November 11 New York summit, the company opened free early access to ElevenAI, a workplace app for meeting notes and follow-up tasks, while previewing persiste…

    Sources: elevenlabs.io, thenextweb.com
  24. Policy

    England proposes safety certification for school AI, with supplier rules still open

    England's Department for Education opened a consultation on November 10 on a safety-certification scheme for education technology and AI products. It is a proposal for education settings in England, not an already-operating certification program o…

    Sources: gov.uk, consult.education.gov.uk
  25. Agents

    Ancep raises $8.5 million for AI-assisted commercial insurance placement

    Ancep announced $8.5 million in funding on November 10 to expand AI tools for commercial insurance brokers. Creandum led the round, with Kleiner Perkins and Vicus Ventures participating. The London company’s legal adviser, Wilson Sonsini, confirme…

    Sources: wsgr.com, ancep.com +1
  26. Policy

    AI’s power race faces a gas lock-in risk, but announced plants are not built capacity

    BloombergNEF’s November 11 report identifies a climate risk in the rush to energize data centers: temporary on-site gas generation can become a durable part of the system. It tracks 181GW of announced gas generation proposed alongside facilities, …

    Sources: about.bnef.com
  27. Models

    Anthropic’s November 12 policy puts hardware safety outside the model

    Anthropic’s revised Usage Policy carries a November 12 effective date, bringing a practical checkpoint for developers connecting Claude to equipment. This is the effective date of a policy announced October 8, not a new model release or a new stat…

    Sources: anthropic.com
  28. Agents

    United restricts unauthorized AI agents across its booking channels

    United Airlines said on November 11 that it is restricting unauthorized third-party personal AI assistants on its digital platforms and through travel agencies and aggregators. The move extends the contest over agent-driven commerce into airline d…

    Sources: united.mediaroom.com, travelweekly.com +1
  29. Models

    Workspace Models move robot memory reasoning into training

    MIT and Carnegie Mellon researchers are studying a way to give robots useful memories without repeatedly asking a large vision-language model what to do next. Their Workspace Models approach is receiving new attention through an MIT account dated …

    Sources: csail.mit.edu, arxiv.org +1
  30. Infrastructure

    NVIDIA puts OpenShell into CNCF and previews more Kubernetes infrastructure

    NVIDIA and the Cloud Native Computing Foundation announced on November 10 that OpenShell, a runtime for controlling autonomous AI agents, is joining CNCF as a Sandbox project. The move places the project within a community governance structure as …

    Sources: cncf.io, blogs.nvidia.com
  31. Agents

    FastMCP turns agent skills into server-delivered artifacts

    FastMCP 4.2.0 lets an MCP server distribute the instructions and files that make up an agent skill. Its client checks file sizes and digests before exporting a skill locally. The experimental skills extension still has limited host support; ordina…

    Sources: github.com, pypi.org
  32. Robotics

    Beijing’s robot plan pairs factory-scale ambitions with lifecycle safety checks

    Beijing has set out a 2026–2028 embodied-robotics action plan that treats manufacturing, training infrastructure and safety governance as parts of the same industrial buildout. The plan was issued on November 11; an official explanation followed o…

    Sources: beijing.gov.cn, jxj.beijing.gov.cn +1
  33. Robotics

    Praxis-1 adds a low-cost robot-arm path, but consumer-GPU validation still matters

    Runway's Praxis-1 repository now lists SO-100/SO-101 checkpoints, including a quantization-aware-trained NVFP4 version. This materially extends the earlier DROID and bimanual release to another robot platform. The repository supplies a joint-space…

    Sources: github.com, jls42.org
  34. Models

    Decision models can be better teachers than actors, a new study finds

    A decision model does not need to be a strong agent to help train one. That is the central result of a November 12 preprint from researchers at Zhejiang University and Baidu, which compares choosing actions with judging completed trajectories. The…

    Sources: arxiv.org, github.com
  35. Robotics

    PRISM gives robot controllers a more explicit way to learn physical interactions

    University of Michigan researchers are highlighting PRISM, a neural-network component designed to improve robot control by learning interactions between physical variables directly. A November 10 university announcement tied the work to the Confer…

    Sources: news.engin.umich.edu, arxiv.org +1
  36. Agents

    Google launches Gemini Decision 1 for the small choices inside AI agents

    Google has introduced Gemini Decision 1 and a dedicated Decisions API for bounded tasks such as choosing a tool, routing a request or grading retrieved context. Its November 11 API changelog confirms the release through the Gemini API. According t…

    Sources: blog.google, aistudio.google.com
  37. Agents

    Claude Code tightens approval and credential boundaries in 2.1.325

    Anthropic's November 13 Claude Code release changes sandboxed Bash to the normal permission flow by default. It also addresses credential selection in background sessions, provider-policy enforcement during token counting, and several approval-rul…

    Sources: github.com
  38. Tools

    Vercel fixes approval replay so validation cannot rewrite the action that was approved

    Vercel’s November 13 AI SDK 7.0.173 release fixes a subtle approval boundary: validating or refining a tool call must not change the arguments a person has already approved. The patch prevents in-place schema validation and input refinement from m…

    Sources: github.com
  39. Policy

    SEBI previews AI guardrails, not a finished rulebook

    India's securities regulator is preparing more detailed AI guidance, with accountability and human oversight at the center of the discussion. That is a policy signal, not evidence that a new rulebook has already taken effect. According to PTI's No…

    Sources: sebi.gov.in, yourstory.com
  40. Agents

    Kindo raises $30 million for the control layer around enterprise agents

    Kindo announced $30 million in new financing on November 12, co-led by Crosspoint Capital and Forgepoint Capital. That brings its cumulative funding to $57.6 million, not a $57.6 million new round. The distinction matters because some headline tre…

    Sources: kindo.ai, forgepointcap.com +2
  41. Agents

    Cloudflare gives Workers a spending brake, not a universal AI budget

    Cloudflare announced Spend Management on November 13, opening a beta for pay-as-you-go customers that starts with Workers and Durable Objects. Customers can configure product-level spending limits and alerts through the dashboard, API or CLI. When…

    Sources: blog.cloudflare.com
  42. Policy

    Runway opens Praxis-1 robot-policy weights and an experimental consumer-GPU path

    Runway made Praxis-1 publicly available on November 11, turning its earlier world-action-model announcement into a release robotics teams can inspect and adapt. The company says the weights, inference code and fine-tuning tools are covered by Apac…

    Sources: runway.com, github.com
  43. Agents

    SGLang's 1.0 plan puts predictable behavior ahead of another performance headline

    Event: November 12, 2026 release tracker opened. Status: roadmap, not a verified 1.0 release. First covered in this run: November 13, 2026, 11:53 UTC. SGLang has outlined a stability-focused path to version 1.0, with explicit configuration errors,…

    Sources: github.com
  44. Policy

    COP31 guidance asks AI companies to show both information quality and environmental costs

    The Global Initiative for Information Integrity on Climate Change launched AI priorities at COP31 on November 13. Brazil, the UN and UNESCO are calling for companies to assess and publish how their systems affect climate information, alongside dis…

    Sources: unfccc.int
  45. Models

    Langfuse adds model-scoped gateway keys and tighter session handling

    Langfuse’s 4.80.0 release introduces gateway-key restrictions based on model patterns, alongside Google AI Studio and Microsoft Azure provider support. Existing gateway keys can have their names, metadata and model selections edited. The interface…

    Sources: github.com
  46. Guth News

    llama.cpp reaches 1.0, making compatibility a first-class promise for local AI

    The most important change in llama.cpp 1.0 is not another benchmark. It is a clearer contract for the developers building local-AI applications on top of it. The maintainers released version 1.0.0 of both llama.cpp and its ggml tensor library on N…

    Sources: huggingface.co, github.com
  47. Agents

    JustRUA improves robot-agent scores by changing the harness, not training a new policy

    A November 11 preprint tests the software surrounding GPT-6 Astra rather than introducing another learned robot policy. The authors report an evaluation score rising from 29.7 to 50.3 in RoboDojo simulation. The retained changes preserve episode h…

    Sources: arxiv.org
  48. Models

    Datasette opens a shorter path from local models to data agents

    Datasette's model-integration plugin has a useful new connection option. The 0.1a11 prerelease supports direct configuration of OpenAI-compatible endpoints without an extra provider plugin. That opens a simpler route to models served on a develope…

    Sources: github.com
  49. Policy

    OpenAI sets community commitments for future campuses it owns and operates

    A new compact promises local reporting and energy-cost protections, alongside separate proposals for industry-wide rules. OpenAI published a National Community Compact and a Compute Campus Policy Blueprint on November 5, setting out commitments fo…

    Sources: openai.com, cdn.openai.com
  50. Policy

    Norway pauses new data-center grid reservations as it prepares a licensing regime

    The temporary grid measures have exceptions, while the government's broader licensing plans are still being developed. Norway is moving to regulate data-center growth through two distinct measures: a temporary pause in new grid-capacity reservatio…

    Sources: statnett.no, regjeringen.no
  51. Guth News

    Microsoft reportedly warns a Chinese optical-transceiver ban could disrupt AI data-center construction

    Industry lobbying targets a proposed restriction, with the scope and timing of any final measure still unsettled. Microsoft has urged the Trump administration to abandon plans restricting Chinese optical-transceiver imports, Reuters reported Novem…

    Sources: investing.com, itic.org +1
  52. Guth News

    Reported H200 stockpile in Hong Kong highlights the gap between chip supply and usable AI capacity

    Reuters says about 600,000 NVIDIA processors await mainland Chinese import clearance, while some smaller shipments have entered. About 600,000 NVIDIA H200 chips were stored in Hong Kong awaiting shipment to mainland China, Reuters reported Novembe…

    Sources: streetinsider.com, media.bis.gov +3
  53. Guth News

    French expert commission urges stronger safeguards for AI assistants and children

    A French expert commission has delivered 18 recommendations aimed at protecting people who use generative AI, placing chatbot design, children's safety and AI-enabled toys under scrutiny. The Élysée published its account of the report on November …

    Sources: elysee.fr, conseil-ia-numerique.fr
  54. Models

    Stanford's third AI100 report examines the costs and choices behind AI's spread

    Stanford HAI highlighted the third One Hundred Year Study on Artificial Intelligence report on November 10, describing a technology whose adoption has expanded faster than governance has adapted. AI Goes Mainstream: The Scaling Era and the Rise of…

    Sources: hai.stanford.edu, ai100.stanford.edu
  55. Business

    Democrats prepare AI oversight agenda as congressional control shifts

    Democrats are preparing legislation on AI safety, corporate liability and electricity costs as they approach control of Congress in January, Reuters reported November 6. The report describes plans to protect consumers from data-center-related powe…

    Sources: reuters.com, trahan.house.gov +2
  56. Business

    AmeriLife expands Palantir partnership after putting AI workflows into production at Crump

    guthlabs.newsroom | Enterprise AI | November 5, 2026 AmeriLife announced an enterprise partnership with Palantir on November 5, extending Foundry and the Artificial Intelligence Platform across its Health and Wealth distribution organizations. The…

    Sources: amerilife.com, palantir.com
  57. Agents

    Google previews Gemini for healthcare, connecting records and administrative work

    guthlabs.newsroom | Healthcare AI | November 10, 2026 Google Cloud introduced Gemini for healthcare on November 10, bringing a specialized agentic offering to clinical, operational and administrative workflows. Launch participants include HCA Heal…

    Sources: googlecloudpresscorner.com, cloud.google.com
  58. Policy

    China outlines comprehensive AI legislation and sector testing plans

    China plans to accelerate comprehensive AI legislation and develop safety standards covering the technology's lifecycle, an economic-planning official said on November 3. Wang Ruomeng, deputy director of the National Development and Reform Commiss…

    Sources: ndrc.gov.cn
  59. Agents

    Kyndryl and ServiceNow bring operational context to autonomous IT workflows

    guthlabs.newsroom | Enterprise AI | November 9, 2026 Kyndryl announced on November 9 that its Autonomous Operations offering for ServiceNow was available through the ServiceNow Store. Co-developed by the companies and powered by Kyndryl Bridge, th…

    Sources: kyndryl.com
  60. Policy

    SpaceX’s proposed $24 billion Starbase power and compute project faces a December tax hearing

    The city has scheduled consideration of incentives, while the proposal leaves generating and computing capacity unspecified. SpaceX is proposing a gas-fired power plant and supercomputing facility in Starbase, Texas, with approximately $24 billion…

    Sources: streetinsider.com, starbase.texas.gov +1
  61. Research

    AP report raises questions about White House ties to US AI safety evaluators

    Staff at the US government's AI evaluation center were asked to report informally to a White House office, according to a November 8 Associated Press report citing two former officials. The account raises questions about the independence of the Ce…

    Sources: abcnews.com, nist.gov
  62. Agents

    Commvault and Zscaler pair AI controls with a plan for recovering from agent mistakes

    guthlabs.newsroom | AI Security | November 10, 2026 Commvault and Zscaler announced a strategic partnership on November 10 aimed at combining controls over enterprise AI use with recovery after harmful agent activity. The companies plan to collabo…

    Sources: commvault.gcs-web.com, ir.commvault.com
  63. Agents

    Vanta agrees to buy Pillar Security, with AI enforcement on a 2027 roadmap

    guthlabs.newsroom | AI Security | November 9, 2026 Vanta announced on November 9 that it had signed a definitive agreement to acquire Pillar Security, adding technology for testing and protecting AI applications and agents. The agreement is a step…

    Sources: vanta.com, pillar.security
  64. Guth News

    Singapore acknowledges the mental workload of checking AI output

    Singapore's Ministry of Manpower says employees who closely monitor and validate AI-generated work can face additional cognitive demands, including mental fatigue. In a November 3 parliamentary response, the ministry said it would continue examini…

    Sources: mom.gov.sg, tal.sg
  65. Tools

    UK guidance limits Northern Ireland AI Act rules to closely integrated physical products

    UK government guidance published November 4 explains a limited application of the EU AI Act in Northern Ireland, focused on AI closely integrated with physical products placed on that market. It sets November 10 as the start for certain provisions…

    Sources: assets.publishing.service.gov.uk
  66. Models

    Claude Enterprise adds security controls, with different limits for hooks, sign-out and monitoring

    Anthropic announced a set of Claude Enterprise security and administration updates on November 5, including a security center, centralized management and generally available inference hooks. The release gives security teams more ways to inspect co…

    Sources: claude.com, support.claude.com
  67. Agents

    WhiteMatter raises $20 million to put policy controls around enterprise AI agents

    guthlabs.newsroom | AI Security | November 10, 2026 WhiteMatter, the AI-security company previously known as QuilrAI, announced a $20 million Series A on November 10. Blumberg Capital led the round, with existing investors Crew Capital and Sprout …

    Sources: prnewswire.com, crew.vc
  68. Agents

    CrowdStrike's Fiddler and Dreadnode deals target the full life of an AI agent

    CrowdStrike announced agreements on November 3 to acquire Fiddler AI and Dreadnode, adding agent observability and automated AI security testing to its planned Falcon offering. The two deals address different questions: whether an agent can be man…

    Sources: ir.crowdstrike.com, crowdstrike.gcs-web.com +1
  69. Models

    Australia plans mandatory frontier AI rules and a dedicated safety regulator

    Australia plans to make frontier AI developers operating in the country subject to mandatory safety obligations and oversight by a new regulator, the government announced on November 10. The proposal would put independent model evaluation, inciden…

    Sources: pm.gov.au, abcnews.com +1
  70. Tools

    Gemini Omni 1.2 Flash lowers video pricing, with workflow limits worth reading

    Google made Gemini Omni 1.2 Flash generally available on November 5, extending native video generation to 15 seconds and lowering the advertised 720p price to $0.07 per output second. Google describes that as a 30% reduction from Omni 1.1. The upd…

    Sources: blog.google, docs.cloud.google.com +2
  71. Models

    Nous releases Arete-27B to let builders change a model’s political framing

    Nous Research has released Arete-27B, a downloadable model designed to change how Alibaba’s Qwen3.8-27B handles political and value-laden questions. The repository history records the initial upload on November 6, and the model card lists an Apach…

    Sources: huggingface.co, nousresearch.com
  72. Models

    Claude Fable and Mythos 5.2 bring lower prices and new migration checks

    Anthropic released Claude Fable 5.2 and Claude Mythos 5.2 on November 9, updating its models for demanding technical work while retaining different access rules. The pair shares the same underlying model. Fable adds cybersecurity and biology safeg…

    Sources: anthropic.com, platform.claude.com +2