# Nieuwsbrief 16 september 2026: OpenAI bevestigt safety-gesprekken met Google en Anthropic

[Naar de inhoud](#lm-inhoud)Netwerk/NL[EN](/en/nieuwsbrief-2026-09-16)[Hubhub.llmnet.nlModellen vergelijken op taak, taal, kosten en licentie.](https://hub.llmnet.nl/)[Communitycommunity.llmnet.nlPrompttechnieken, patronen en systeemprompts.](https://community.llmnet.nl/)[APIapi.llmnet.nlLLM's robuust in software: rate limits, routing, structured output.](https://api.llmnet.nl/)[Consultancyconsultancy.llmnet.nlAI invoeren in een organisatie, van pilot tot productie.](https://consultancy.llmnet.nl/)[Nieuwsnieuws.llmnet.nlOntwikkelingen in AI, geduid voor Nederland.](https://nieuws.llmnet.nl/)[Benchmarkbenchmark.llmnet.nlZelf meten wat AI-kwaliteit is, voor jouw taken.](https://benchmark.llmnet.nl/)[Vacaturesvacatures.llmnet.nlAI-rollen, salarissen en carrièrepaden in Nederland.](https://vacatures.llmnet.nl/)[Lerenleren.llmnet.nlAI-concepten in gewoon Nederlands, van beginner tot bouwer.](https://leren.llmnet.nl/)[Gidsgids.llmnet.nlAI privé draaien op eigen Mac, pc, NAS of thuisserver.](https://gids.llmnet.nl/)[Directorydirectory.llmnet.nlHet AI-ecosysteem in kaart: tools, modellen, bedrijven.](https://directory.llmnet.nl/)[Radarradar.llmnet.nlSignalen uit X, onderzoek en communities voor indie developers.](https://radar.llmnet.nl/)[Appsapps.llmnet.nlReviews van AI-apps en open-source repo's, met tips voor wie zelf bouwt.](https://apps.llmnet.nl/)[Trendingtrending.llmnet.nlElke dag de AI-trends van X, geduid in het Nederlands.](https://trending.llmnet.nl/)[llmnet.nl — hoofdsite](https://llmnet.nl/)[](https://x.com/intent/post?url=https%3A%2F%2Fapps.llmnet.nl%2Fnieuwsbrief-2026-09-16&text=Nieuwsbrief%2016%20september%202026%3A%20OpenAI%20bevestigt%20safety-gesprekken%20met%20Google%20en%20Anthropic)[](https://www.linkedin.com/sharing/share-offsite/?url=https%3A%2F%2Fapps.llmnet.nl%2Fnieuwsbrief-2026-09-16)[](https://www.reddit.com/submit?url=https%3A%2F%2Fapps.llmnet.nl%2Fnieuwsbrief-2026-09-16&title=Nieuwsbrief%2016%20september%202026%3A%20OpenAI%20bevestigt%20safety-gesprekken%20met%20Google%20en%20Anthropic)[](#)

[← Alle nieuwsbrieven](/nieuwsbrieven)

# Nieuwsbrief woensdag 16 september 2026

Dagelijkse update over de AI-tools die we volgen. Koppen komen van publieke bronnen (X, GitHub, Hacker News); de Nederlandse duiding en samenvatting zijn machinaal geschreven. Elke kop leidt naar onze eigen nieuwspagina over dat onderwerp; de bronlink staat ernaast.

## In het kort

Een OpenAI-woordvoerder bevestigt tegenover CNBC gesprekken met Anthropic en Google over een AI-veiligheidsorgaan. Cloudflare maakt het mogelijk om toegang tot individuele Workers af te scherpen met smallere Developer Platform-rollen. Claude Code verschijnt als 2.1.273 met gateway hint-headers en een MCP-disconnect-notificatie, en Trail of Bits brengt drie skills uit voor patch-validatie. Shield AI onderhandelt over een ronde die het defensiebedrijf op minstens 20 miljard dollar zou waarderen.

## Claude Code & agents

- [feat(models): OpenAI and Anthropic support via per-node model selection](/nieuws-claude-agents#i-1be4820683) · [bron: GitHub](https://github.com/himanshu-nakrani/aegis/pull/51) · 2026-09-14

- [I delegated a one line edit to a Claude Code subagent. It cost 77,000 tokens. So I measured what subagents actually cost, where those tokens](/claude-code-nieuws#i-acd9f52ac9) · [bron: X](https://x.com/ucsandman/status/2099900535499354148) · 2026-09-15

- [Show HN: Bough, the agent I built to replace Claude Code at work](/claude-code-nieuws#i-1b45f71934) · [bron: Hacker News](https://github.com/andreylukin/bough) · 2026-09-15

- [Coordinating multiple coding agents is harder than asking one to write code. Ruflo is an agent meta-harness for builders using Claude Code a](/claude-code-nieuws#i-a5d17d7bc6) · [bron: X](https://x.com/DanKornas/status/2099520807563120885) · 2026-09-14

## Cloudflare OS & self-hosted

- [Story 4.19: Architectural Spike — Evaluate Cloudflare Browser Services vs. GitHub Actions CI for Playwright UI & Integration Testing](/nieuws-self-hosted#i-ac26ff2fd8) · [bron: GitHub](https://github.com/jacobmiller22/chrishop/issues/206) · 2026-09-12

- [You can now scope access to individual Workers and assign narrower Developer Platform roles, so teammates, CI tokens, and agents get only th](/nieuws-self-hosted#i-9ce67039d4) · [bron: X](https://x.com/Cloudflare/status/2099846955887149439) · 2026-09-15

- [feat(cloudflare): support outbound workers in Browser Run](/cloudflare-os-nieuws#i-3128fea361) · [bron: GitHub](https://github.com/cloudflare/playwright/pull/228) · 2026-09-15

- [Built a self-hosted homelab on an old laptop — sanity checking my setup (security, Docker, media stack)](/nieuws-self-hosted#i-7b450fd99e) · [bron: Reddit](https://www.reddit.com/r/homelab/comments/1wg1u93/built_a_selfhosted_homelab_on_an_old_laptop/) · 2026-09-14

## Agent-tooling & MCP

- [Cardiovascular Imaging AI Startup Corologic Raises Seed Funding from Kakao Ventures and Seoul Techno Holdings https://en.wowtale.net/2026/09](/nieuws-mcp-agent-tooling#i-eed4b330b7) · [bron: X](https://x.com/audiumcp/status/2099877482811814034) · 2026-09-15

- [A new WhatsApp Business MCP server lets developers use AI coding agents like Claude, Cursor, Codex, and ChatGPT to handle setup, messaging t](/codex-nieuws#i-c89cd6929b) · [bron: X](https://x.com/TechCrunch/status/2099955188467405227) · 2026-09-15

## Modellen & prijzen

- [Building your first LLM API call in Python (step by step)](/nieuws-modellen-prijzen#i-c97c118da1) · [bron: Hacker News](https://heymeraki.substack.com/p/aie_10-building-it) · 2026-09-15

- [ok everyone, i took one for the team here's the data we all wanted to see - real token value of each LLM subscription, empirically measured](/nieuws-modellen-prijzen#i-61a80b5cc8) · [bron: X](https://x.com/kunchenguid/status/2098256018836963382) · 2026-09-11

- [Show HN: Determinstic LLM inference for lowest price Gemma 4, with Windows XP](/nieuws-modellen-prijzen#i-feb336d614) · [bron: Hacker News](https://www.tokendelivery.ai/) · 2026-09-12

- [Admin: model configuration per stage, and what each stage has spent — needs a design session](/nieuws-modellen-prijzen#i-de2b738b06) · [bron: GitHub](https://github.com/tadasant/motet/issues/92) · 2026-09-13

## Skills & plugins

- [We found 5 open source AI github repos to level up your AI research and building skills. 1. LLM-FINGERPRINT-DETECTOR This tool fingerprints](/nieuws-skills-plugins#i-a61c1ea4b2) · [bron: X](https://x.com/sentient_found/status/2099498411460612287) · 2026-09-14

- [We open-sourced 73 business Agent Skills — SEO/AEO, marketing, copy, ops, market sizing, exec advisory. Not script wrappers: each is ~2KB of](/nieuws-skills-plugins#i-6f31b376ac) · [bron: X](https://x.com/meshcodeai/status/2098790788734001395) · 2026-09-12

- [feat: add AI Agent Skills & Plugins section (bootgs/skills)](/nieuws-skills-plugins#i-149b394d62) · [bron: GitHub](https://github.com/oshliaer/google-apps-script-awesome-list/pull/209) · 2026-09-13

- [.@trailofbits released 3 skills to help write patches: `post-patch-validation` helps agents test that fixes are correct https://t.co/wtsGX8Q](/nieuws-skills-plugins#i-c50a41a58d) · [bron: X](https://x.com/dguido/status/2099865541380845848) · 2026-09-15

## Repo's, releases & updates

- [[RESEARCH] AI research startup Listen Labs scrubbed a $1.5B funding round for Salesforce talks - TechCrunch](/nieuws-repos-releases#i-0d915657ce) · [bron: GitHub](https://github.com/fratei/creative-ware-hq/issues/1094) · 2026-09-16

- [Slack: status indicator silently broken with slack-sdk >= 3.44 (agents.sessions.setStatus rejects free-text and empty status)](/hermes-nieuws#i-f95b949f9f) · [bron: GitHub](https://github.com/NousResearch/hermes-agent/issues/110374) · 2026-09-13

- [fix(antigravity): stop health checks from filling the disk with _MEI folders](/nieuws-repos-releases#i-6e03d9e082) · [bron: GitHub](https://github.com/pingdotgg/t3code/pull/12008) · 2026-09-16

- [Open weights are not open source: Why AI's favorite label is under dispute](/nieuws-repos-releases#i-fa2ebb7707) · [bron: Hacker News](https://www.theregister.com/columnists/2026/09/15/open-weights-are-not-open-source-why-ais-favorite-label-is-under-dispute/5295436) · 2026-09-15

## Financiering & markt

- [Shield AI is in talks for a funding round that would value the defense startup at at least $20 billion, following recent Air Force and Penta](/nieuws-financiering-markt#i-13fc5371fc) · [bron: X](https://x.com/theinformation/status/2099845801727983987) · 2026-09-15

- [Israeli AI construction startup Buildots raises $130 million to fuel global expansion New funding values Buildots’ push to bring AI oversigh](/nieuws-financiering-markt#i-4ca8fe35a9) · [bron: X](https://x.com/ynetnews/status/2099570921119863086) · 2026-09-14

- [AI Startup/Australia](/nieuws-financiering-markt#i-d8d5f4645f) · [bron: Reddit](https://www.reddit.com/r/cofounderhunt/comments/1wglfch/ai_startupaustralia/) · 2026-09-15

- [This NVIDIA program for AI startups gives you free cloud credits, technical training, and VC intros - all for zero equity: It’s called NVIDI](/nieuws-financiering-markt#i-5ae041edf2) · [bron: X](https://x.com/wecraveai/status/2099903980063379625) · 2026-09-15

## Achtergrond: wat er verder speelde

- Claude Code 2.1.272 → 2.1.273 overnight: npm latest/next = 2.1.273 (npm-tijd 15 sept 18:06 UTC); GitHub v2.1.273 (~15 sept 20:22 UTC). Nieuw o.a. gateway hint-headers, MCP-disconnect-notificatie + /mcp, forken van remote-control-sessies vanaf de Claude-app als background op je machine, managed-MCP-settings-fix, Bedrock/Vertex/Foundry 401/403-messaging, context-meter/auto-compact overcount-fix, macOS Read/screenshot-pad-fix. Dist-tag stable is nu 2.1.267 (was gisteren 2.1.236 — tag-promotie; package zelf dateert van 9 sept). ([1](https://github.com/anthropics/claude-code/releases/tag/v2.1.273), [2](https://www.npmjs.com/package/@anthropic-ai/claude-code), [3](https://github.com/anthropics/claude-code/releases/tag/v2.1.267))

- Antigravity CLI 1.2.4 (16 sept ~03:54 UTC): /skills reload, /model autocomplete + fuzzy picker, betere OAuth-scope-fouten, plus MCP schema-/subagent-/hooks-/resume-fixes. ([4](https://github.com/google-antigravity/antigravity-cli/releases/tag/1.2.4))

- Codex alpha → 0.155.0-alpha.10 (16 sept ~04:26 UTC); stable blijft 0.154.0. ([7](https://www.npmjs.com/package/@openai/codex))

- @ai-sdk/mcp 2.0.50 (15 sept 17:25 UTC); MCP SDK TS blijft 1.30.0. ([8](https://www.npmjs.com/package/@ai-sdk/mcp), [9](https://www.npmjs.com/package/@modelcontextprotocol/sdk))

- Standards-body (15 sept): CNBC — OpenAI-woordvoerder bevestigt gesprekken met Anthropic en Google over safety-samenwerking / standards-body (Hassabis juli FINRA-achtig voorstel; volgt Amodei pacing). Bloomberg/Straits Times echo. ([10](https://www.cnbc.com/2026/09/15/open-ai-google-anthropic-safety.html), [11](https://www.bloomberg.com/news/articles/2026-09-15/openai-says-it-s-working-with-anthropic-google-on-ai-safety), [12](https://www.straitstimes.com/world/united-states/openai-anthropic-and-google-are-working-to-create-an-ai-standards-body))

- Stil: Hermes v2026.9.14 / 0.21.3; Cursor desktop tracker top 3.20.21 (13 sept); OpenClaw 2026.9.4 (docs 9.5/9.6 = 404); x.ai/news geen Grok 4.7. ([13](https://github.com/NousResearch/hermes-agent/releases/tag/v2026.9.14), [14](https://github.com/oslook/cursor-ai-downloads/blob/main/version-history.json), [15](https://docs.openclaw.ai/releases/2026.9.4), [16](https://x.ai/news))

[Trends op X van deze dag](https://trending.llmnet.nl/trends-2026-09-16) · [← 15 september 2026](/nieuwsbrief-2026-09-15) · [17 september 2026 →](/nieuwsbrief-2026-09-17) · [Archief](/nieuwsbrieven)