ingest(xpost): Brian Roemmele Amazon-Jailbreak Fable 5 als Auslöser der Exportkontrollen
This commit is contained in:
Hector Bot
parent
f863d48d64
commit
2328b1c8ff
5 changed files with 93 additions and 7 deletions
73
raw/xpost/2026-06-13_roemmele-amazon-jailbreak-fable5.md
Normal file
73
raw/xpost/2026-06-13_roemmele-amazon-jailbreak-fable5.md
Normal file
|
|
@ -0,0 +1,73 @@
|
|||
---
|
||||
type: xpost
|
||||
source_url: https://x.com/BrianRoemmele/status/2065835859073741112
|
||||
retrieved: 2026-06-13
|
||||
author: "@BrianRoemmele"
|
||||
is_thread: true
|
||||
quote_count: 2
|
||||
tags: [anthropic, fable5, mythos, jailbreak, amazon, export-controls]
|
||||
---
|
||||
|
||||
# Brian Roemmele: Amazon-Jailbreak von Fable 5 als Auslöser der Exportkontrollen
|
||||
|
||||
## Thread
|
||||
|
||||
### Post 1: "Anthropic Just Shot Itself in the Foot"
|
||||
**URL:** https://x.com/BrianRoemmele/status/2065799324790088005
|
||||
**Posted:** 2026-06-13 14:11:54 GMT
|
||||
**Likes:** 378 | **Retweets:** 83
|
||||
|
||||
> Anthropic launched Fable 5 and Mythos 5, then watched the US government shut them down three days later. The same government their CEO Dario Amodei has been begging for years to regulate AI harder. Now he got exactly what he asked for.
|
||||
>
|
||||
> This is straight-up leadership failure. Dario spent all that time pushing for rules and oversight. Those rules just killed his flagship models overnight. Customers in the middle of builds got cut off. Security teams using the models to find vulnerabilities suddenly had nothing. The company tried to call it a narrow export control thing over a jailbreak, but nobody is buying that spin.
|
||||
>
|
||||
> I helped move big clients off Anthropic the same night. One account alone was worth millions a month. They switched to local open-source models and they are not coming back.
|
||||
>
|
||||
> This is going to leave permanent damage. Customer exodus, key people leaving, and their IPO plans looking dead by the end of summer.
|
||||
>
|
||||
> This hurts US AI competitiveness and national security work. It pushes people toward open-source options, including ones from China.
|
||||
>
|
||||
> All because Anthropic positioned itself as the "safe and responsible" company that wanted government help. Now that help just flipped the off switch on their best stuff.
|
||||
>
|
||||
> [... Dario Amodei's history of fear-mongering and delay tactics ...]
|
||||
>
|
||||
> In two years, everyone will have Mythos-class AI — or better — running in their pocket, on their devices, with no guardrails, no corporate nanny filters, and no remote kill switch.
|
||||
>
|
||||
> Local, open-source, unstoppable. History is going to laugh at this entire episode: the CEO who spent years slowing everyone down only to watch his own company self-destruct by inviting the regulators to the party.
|
||||
>
|
||||
> Dario wanted regulation. He got it. The rest of the industry gets the lesson: inviting the state into your tech is a fast way to lose control of it.
|
||||
>
|
||||
> Centralized models like this are too fragile.
|
||||
>
|
||||
> Open-source and local alternatives just picked up a lot more users who will never trust a company like Anthropic again.
|
||||
>
|
||||
> This whole mess was completely avoidable. Hubris dressed up as safety advocacy.
|
||||
>
|
||||
> Now the bill is due.
|
||||
|
||||
### Post 2: "A self inflicted wound"
|
||||
**URL:** https://x.com/BrianRoemmele/status/2065823846624850378
|
||||
**Posted:** 2026-06-13 15:49:20 GMT
|
||||
**Likes:** 24 | **Retweets:** 1
|
||||
|
||||
> A self inflicted wound.
|
||||
|
||||
## Kontext (aus Kais Nachricht in OME-Gruppe "Krallenpolitik")
|
||||
|
||||
Brian Roemmele berichtet, dass Amazon-Forscher Fable 5 gejailbreakt haben (Prompt-Injection, um Sicherheitsvorkehrungen zu umgehen) und das Modell dazu gebracht haben, Details zu echten Sicherheitslücken preiszugeben. Statt verantwortungsvoller Offenlegung gegenüber Anthropic haben sie die Ergebnisse an die US-Regierung weitergegeben. Dies war der Auslöser für Commerce Secretary Howard Lutnicks Exportkontrollen gegen Mythos 5 und Fable 5.
|
||||
|
||||
## Bestätigung durch Medienberichte
|
||||
|
||||
Laut Wall Street Journal (WSJ) und Times of India wurde die Jailbreak-Technik von **Amazon-Forschern** getestet. Katie Moussouris, CEO der Cybersicherheitsfirma Luta Security, bestätigte, dass die Forscher eine Reihe von Prompts verwendeten, um das Modell zur Identifizierung von Software-Sicherheitslücken zu bringen.
|
||||
|
||||
Anthropic selbst bestätigte: "We reviewed a demonstration of this specific technique being used to identify a small number of previously known, minor vulnerabilities. These vulnerabilities all appear relatively simple, and we have found that other publicly available models are able to discover them as well without requiring a bypass."
|
||||
|
||||
## Kernpunkte
|
||||
|
||||
- **Amazon** ist Milliarden-Investor in Anthropic und hostet deren Modelle auf AWS
|
||||
- Amazon-Forscher haben Fable 5 gejailbreakt (Prompt-Injection)
|
||||
- Statt an Anthropic gemeldet, an US-Regierung weitergegeben
|
||||
- Das war der konkrete Auslöser für die Exportkontrollen (nicht nur "ein anderes Unternehmen")
|
||||
- Anthropic reagierte mit 700-Wörter-Statement und behauptete "begrenzter, nicht universeller Jailbreak"
|
||||
- Ironie: Anthropic hat jahrelang Sicherheitsfrömmigkeit zelebriert, wird jetzt vom eigenen Investor vorgeführt
|
||||
- Exportkontrollen = Modelle tot, Nutzer gestrandet, IPO-Aussichten getrübt
|
||||
|
|
@ -1,7 +1,8 @@
|
|||
---
|
||||
created: 2026-06-05
|
||||
updated: 2026-06-13
|
||||
sources: [blog/2026-06-04_anthropic-fordert-ki-pause.md, blog/2026-05-15_ai-news-roundup-may-2026.md, blog/2026-06-13_trump-export-controls-anthropic-mythos-fable.md, blog/2026-06-13_malone-biological-ai-biosecurity.md]
|
||||
sources: [blog/2026-06-04_anthropic-fordert-ki-pause.md, blog/2026-05-15_ai-news-roundup-may-2026.md, blog/2026-06-13_trump-export-controls-anthropic-mythos-fable.md, blog/2026-06-13_malone-biological-ai-biosecurity.md, xpost/2026-06-13_roemmele-amazon-jailbreak-fable5.md]
|
||||
sources: [blog/2026-06-04_anthropic-fordert-ki-pause.md, blog/2026-05-15_ai-news-roundup-may-2026.md, blog/2026-06-13_trump-export-controls-anthropic-mythos-fable.md, blog/2026-06-13_malone-biological-ai-biosecurity.md, xpost/2026-06-13_roemmele-amazon-jailbreak-fable5.md]
|
||||
tags: [concept, regulation, safety, government, export-controls, anthropic]
|
||||
---
|
||||
|
||||
|
|
@ -22,7 +23,7 @@ Ab Mai 2026: Microsoft, Google DeepMind und xAI verpflichten sich zu Pre-Release
|
|||
Am **12. Juni 2026** verhängte die Trump-Administration Exportkontrollen gegen Anthropics Spitzenmodelle **Mythos 5 und Fable 5**:
|
||||
|
||||
- **Was:** Commerce Secretary Howard Lutnick ordnete an, dass beide Modelle nicht mehr an ausländische Regierungen, Unternehmen oder Personen (auch innerhalb der USA) lizenziert werden dürfen
|
||||
- **Auslöser:** Ein anderes Unternehmen behauptete, Mythos jailbreaken zu können — das Weiße Haus sah nationale Sicherheitsrisiken
|
||||
- **Auslöser:** Amazon-Forscher jailbreakten Fable 5 per Prompt-Injection und gaben die Ergebnisse an die US-Regierung weiter (nicht an Anthropic). Laut WSJ und Times of India testeten Amazon-Forscher die Technik, um das Modell zur Identifizierung von Software-Sicherheitslücken zu bringen. Katie Moussouris (Luta Security) bestätigte die Amazon-Verbindung. Das Weiße Haus sah nationale Sicherheitsrisiken.
|
||||
- **Anthropics Reaktion:** Komplette Abschaltung beider Modelle für alle Kunden („we believe this is a misunderstanding")
|
||||
- **Dauer:** Voraussichtlich mehrere Wochen — über den 22. Juni hinaus (geplanter Wechsel von Free zu Paid bei Fable 5)
|
||||
- **Anthropics Warnung:** Der Standard würde „im Wesentlichen alle neuen Modelldeployments aller Frontier-Modellanbieter stoppen"
|
||||
|
|
|
|||
|
|
@ -2,7 +2,7 @@
|
|||
|
||||
*Auto-generated: 2026-06-13*
|
||||
|
||||
*Letzte Aktualisierung: 2026-06-13*
|
||||
*Letzte Aktualisierung: 2026-06-13 (2. Update)*
|
||||
|
||||
## Architecture
|
||||
|
||||
|
|
@ -19,7 +19,7 @@
|
|||
|
||||
| Seite | Beschreibung | Quellen |
|
||||
|-------|-------------|---------|
|
||||
| [Anthropic Claude](tools/anthropic-claude.md) | Claude Opus 4.7, Mythos, Business-Kennzahlen, Pentagon-Streit | 2 |
|
||||
| [Anthropic Claude](tools/anthropic-claude.md) | Claude Opus 4.7, Mythos, Business-Kennzahlen, Pentagon-Streit, Amazon-Jailbreak | 4 |
|
||||
| [OpenAI GPT Models](tools/openai-gpt.md) | GPT-5.5, GPT-5.5 Instant, Pricing | 1 |
|
||||
| [Ecosystem Tools (April 2026)](tools/ecosystem-tools-april-2026.md) | GBrain, SearXNG, Hermes Atlas, X-API, Claude Shortcuts | context-tree |
|
||||
| [Kimi K2.7 Code](tools/kimi-k2.7-code.md) | Coding-fokussiertes agentisches Modell (Moonshot AI) auf Ollama Cloud | other/2026-06-13_kimi-k2.7-code-ollama.md |
|
||||
|
|
@ -29,7 +29,7 @@
|
|||
| Seite | Beschreibung | Quellen |
|
||||
|-------|-------------|---------|
|
||||
| [LLM Knowledge Base](concepts/llm-knowledge-base.md) | Grundlage des RamaDama Wiki (Karpathy-Pattern) | 0 |
|
||||
| [AI Regulation 2026](concepts/ai-regulation-2026.md) | Government Reviews, Anthropic-Pause-Forderung, Mythos-Kontroverse, Trump-Exportkontrollen | 4 |
|
||||
| [AI Regulation 2026](concepts/ai-regulation-2026.md) | Government Reviews, Anthropic-Pause-Forderung, Mythos-Kontroverse, Trump-Exportkontrollen, Amazon-Jailbreak | 5 |
|
||||
| [Biologische KI & Biosecurity-Staat](concepts/ai-biological-biosecurity.md) | Malone-Kritik: Machtkonzentration unter Biosecurity-Deckmantel, GOF-Kontroversen, DeepSeek-Paradoxon | 1 |
|
||||
| [AI Agents 2026](concepts/ai-agents-2026.md) | Wandel zu autonomen Agenten, Enterprise-Adoption, Infrastruktur | 1 |
|
||||
| [Subconscious Agent v2.1](concepts/subconscious-agent.md) | Hard Synthesis, Execution Gap, bekannter Bug, Outcomes | raw/other/2026-06-07_subconscious-runner-script.md |
|
||||
|
|
@ -58,6 +58,7 @@
|
|||
| `raw/blog/2026-06-05_ai-news-roundup-may-2026.md` | blog | Latest AI & Technology News Roundup – May 2026 |
|
||||
| `raw/other/2026-06-07_subconscious-runner-script.md` | other | Subconscious Agent Evidence Runner Script v2.1 |
|
||||
| `raw/podcast/ome20-special-ufologie-physik-bewusstsein-2026-06-12.md` | podcast | OME20 (Special ed.) — Ufologie, Physik und Bewusstsein |
|
||||
| `raw/xpost/2026-06-13_roemmele-amazon-jailbreak-fable5.md` | xpost | Brian Roemmele: Amazon-Jailbreak von Fable 5 als Auslöser der Exportkontrollen |
|
||||
| `raw/blog/2026-06-13_trump-export-controls-anthropic-mythos-fable.md` | blog | Trump admin blocks foreign access to Anthropic's Mythos 5 and Fable 5 |
|
||||
| `raw/blog/2026-06-13_malone-biological-ai-biosecurity.md` | blog | Malone: The AI They Don't Want You to Have — Biosecurity als Machtkonzentration |
|
||||
| `raw/xpost/2026-06-11_colgate-llm-purchase-intent-ssr.md` | xpost | Colgate: LLMs Predict Purchase Intent at 90% via SSR |
|
||||
|
|
|
|||
10
wiki/log.md
10
wiki/log.md
|
|
@ -96,6 +96,16 @@
|
|||
- index: updated (neuer raw-Eintrag, aktualisierte Concept-Beschreibung)
|
||||
- log: updated
|
||||
|
||||
## [2026-06-13] Ingest | Brian Roemmele: Amazon-Jailbreak von Fable 5 als Auslöser der Exportkontrollen
|
||||
**Type:** ingest | **Scope:** raw/xpost, wiki/concepts, wiki/tools
|
||||
**Source:** Kai (@PWeber) in OME-Gruppe "Krallenpolitik"
|
||||
**Actions:**
|
||||
- raw: `raw/xpost/2026-06-13_roemmele-amazon-jailbreak-fable5.md` (created — Brian Roemmele X-Thread: Amazon-Forscher jailbreakten Fable 5 per Prompt-Injection, gaben Ergebnisse an US-Regierung statt an Anthropic)
|
||||
- wiki: `concepts/ai-regulation-2026.md` (updated — Amazon-Jailbreak als konkreter Auslöser der Exportkontrollen, WSJ/Times of India-Bestätigung)
|
||||
- wiki: `tools/anthropic-claude.md` (updated — Trigger-Detail: Amazon-Forscher statt "ein anderes Unternehmen")
|
||||
- index: updated (neuer Raw Sources-Eintrag, aktualisierte Quellen-Zähler)
|
||||
- log: updated
|
||||
|
||||
## [2026-06-13] Ingest | Kimi K2.7 Code auf Ollama Cloud (NVIDIA B300)
|
||||
**Type:** ingest | **Scope:** raw/other, wiki/tools
|
||||
**Source:** Pit Weber (@PWeber) in OME News
|
||||
|
|
|
|||
|
|
@ -1,7 +1,8 @@
|
|||
---
|
||||
created: 2026-06-05
|
||||
updated: 2026-06-13
|
||||
sources: [blog/2026-06-04_anthropic-fordert-ki-pause.md, blog/2026-05-15_ai-news-roundup-may-2026.md, blog/2026-06-13_trump-export-controls-anthropic-mythos-fable.md]
|
||||
sources: [blog/2026-06-04_anthropic-fordert-ki-pause.md, blog/2026-05-15_ai-news-roundup-may-2026.md, blog/2026-06-13_trump-export-controls-anthropic-mythos-fable.md, xpost/2026-06-13_roemmele-amazon-jailbreak-fable5.md]
|
||||
sources: [blog/2026-06-04_anthropic-fordert-ki-pause.md, blog/2026-05-15_ai-news-roundup-may-2026.md, blog/2026-06-13_trump-export-controls-anthropic-mythos-fable.md, xpost/2026-06-13_roemmele-amazon-jailbreak-fable5.md]
|
||||
tags: [tool, anthropic, claude, model]
|
||||
---
|
||||
|
||||
|
|
@ -32,7 +33,7 @@ tags: [tool, anthropic, claude, model]
|
|||
Am 12. Juni 2026 verhängte die Trump-Administration Exportkontrollen:
|
||||
- **Mythos 5:** Ursprünglich für vetted Organisationen via Project Glasswing (Cybersicherheitsfokus)
|
||||
- **Fable 5:** Public-Launch am 9. Juni 2026 — Free-Access-Version ohne die Cybersicherheits-Safeguards
|
||||
- **Trigger:** Ein anderes Unternehmen behauptete, Mythos jailbreaken zu können
|
||||
- **Trigger:** Amazon-Forscher jailbreakten Fable 5 per Prompt-Injection und gaben die Ergebnisse an die US-Regierung weiter (WSJ/Times of India). Katie Moussouris (Luta Security) bestätigte die Amazon-Verbindung.
|
||||
- **Anthropics Reaktion:** Komplette Abschaltung beider Modelle für alle Kunden
|
||||
- **Weitere Modelle** (Claude Opus 4.8, ältere Modelle) nicht betroffen
|
||||
|
||||
|
|
|
|||
Loading…
Add table
Reference in a new issue