How AVALW News Works
AVALW News is a global AI-powered newsroom that creates original journalism by synthesizing facts from multiple verified sources. We are not a news aggregator, not a translation service, and not a content scraper. Every article we publish is an original, transformative work created through a structured editorial pipeline.
Our system indexes 437+ public RSS feeds from 59 countries, processes them through AI-driven verification and synthesis, and publishes original articles with full source attribution.
Discovery
The system continuously indexes publicly available RSS feeds from verified news sources worldwide, refreshing every five minutes. We respect robots.txt directives and publisher opt-out requests at all times. We access only public RSS metadata -- never paywalled content, never full article text.
Clustering
Incoming articles are grouped by topic and event. When multiple sources across different countries report the same story, they are merged into a single cluster for cross-referencing and verification. Duplicate and near-duplicate content is identified and grouped.
Source Verification
From each cluster, the system extracts structured facts: WHO is involved, WHAT happened, WHERE, WHEN, WHY, and HOW. A minimum of 5 independent sources must confirm a core fact before it is included in the final article. Unverified claims, single-source allegations, and unattributed speculation are excluded. This multi-source verification requirement is the foundation of our editorial integrity.
Consensus Engine
Our Consensus Engine analyzes source agreement across the cluster: how many sources report the same facts, whether there are contradictions, and whether reporting differs by region. It detects potential bias, identifies gaps in coverage, and produces a reliability assessment for each story. This gives readers a transparent view of how well-supported and widely confirmed each piece of information is.
AI Synthesis
Using Qwen language models running on infrastructure owned by Avalw S.R.L. in the EU, the system generates a completely original article based on verified facts. Zero sentences are copied from any source. The output has its own structure, phrasing, and narrative arc, following professional journalism standards. Every article includes full source attribution with links to every original source used.
AI Image Generation
Stable Diffusion generates unique illustrative images on local GPU hardware owned by Avalw S.R.L. No photographs are taken from any source. All images are abstract and illustrative -- they do not depict identifiable human faces or claim to be photographs of real events. Every image is watermarked "AVALW AI image" and includes machine-readable metadata identifying it as AI-generated synthetic content.
AVALW Live Protocol
In addition to written source synthesis, AVALW News operates a real-time broadcast intelligence system called AVALW Live Protocol. This system functions as an autonomous newsroom with AI agents acting as live reporters monitoring international television news broadcasts around the clock.
Live Broadcast Monitoring
The system simultaneously monitors multiple international television news channels broadcasting publicly and freely available streams. Channels are monitored across multiple languages including English, French, Spanish, Romanian, and others. The system captures audio from these public broadcasts continuously, 24 hours a day, 7 days a week.
Real-Time Speech Transcription
Audio from each channel is transcribed in real time using Whisper, an open-source speech recognition model running locally on infrastructure owned by Avalw S.R.L. No audio is stored permanently. No audio is sent to any third-party service. Transcription happens entirely on-device, on EU-based hardware. The system processes multiple channels simultaneously, with each channel transcribed independently.
Story Detection and Separation
AI agents analyze the combined transcripts from all monitored channels and identify individual news stories. The system distinguishes between actual news reporting and non-news content such as advertisements, channel promotions, music interludes, and presenter chat. Each distinct news event is separated into an independent story for processing. When the same event is discussed on multiple channels, the AI cross-references facts across all sources.
Original Article Generation
For each identified story, the system generates a completely original article. The AI writes as a primary source, as it has directly monitored the live broadcasts. Direct quotes from officials, leaders, and public figures are preserved exactly as spoken. No content is copied from any written publication. The article includes context, analysis, and implications, written to the standards of international wire services.
Editorial Review
Every Live Protocol article passes through an independent AI review system (Qwen 3.5 with advanced reasoning) that verifies factual accuracy against the original transcripts, corrects any name misspellings from audio transcription, ensures the title is original, and checks grammar and journalistic quality. Articles that fail review are not published.
Publication
Approved articles are published with a distinctive "AVALW Live Protocol" label, making it transparent to readers that the article was produced through live broadcast monitoring rather than written source synthesis. An AI-generated illustration is created for each article. The article is deployed globally within minutes of the original broadcast.
Legal Basis for Live Broadcast Monitoring
AVALW Live Protocol monitors only publicly available, freely broadcast television signals accessible without subscription, authentication, or circumvention of any access control. The system listens to what is said publicly on air, the same information available to any viewer worldwide. AVALW does not record, store, or redistribute any audio or video content. Only factual information extracted from public statements is used to create original journalistic works. This practice is protected under:
- Freedom of the press and the right to report on public statements made through mass media
- Fair use / fair dealing principles in multiple jurisdictions, as the output is transformative original journalism
- Berne Convention, Article 2(8): Facts and news of the day are not subject to copyright protection
- EU Copyright Directive: The use of facts from public broadcasts for journalistic purposes is permitted
- Public interest: Making global news accessible in real time serves the public interest in informed democratic participation
Quality Control
Every article published on AVALW News passes through a triple verification process before publication.
- Source verification: Minimum 5 independent sources must confirm core facts
- Consensus check: The Consensus Engine evaluates source agreement and flags contradictions
- Automated quality review: Content is checked for coherence, factual consistency, and compliance with editorial standards
- Continuous editorial oversight is maintained by qualified staff
- Articles that fail quality checks are not published
AI Models Used
All AI processing runs on infrastructure owned by Avalw S.R.L. in the European Union. No user data or content is sent to external AI providers. All models are open-source.
- Text synthesis
- Qwen language models (open-source), hosted locally on AVALW infrastructure in the EU
- Image generation
- Stable Diffusion (open-source), running on local GPU hardware
- Consensus Engine
- Qwen language models with specialized analysis prompts for multi-source verification, bias detection, and reliability assessment
- Data processing
- All models run entirely on servers owned by Avalw S.R.L. -- no cloud AI APIs, no external model providers
EU AI Act Compliance
AVALW News complies with Regulation (EU) 2024/1689 (the EU Artificial Intelligence Act), including Article 50 transparency obligations effective August 2, 2026.
Content Labeling
Every AI-generated article carries a visible disclosure badge stating it was created by artificial intelligence. Every AI-generated image is watermarked "AVALW AI image" and contains machine-readable metadata identifying it as synthetic content.
Model Transparency
We disclose which AI models are used (Qwen for text, Stable Diffusion for images), that they are open-source, and that all processing occurs on our own infrastructure within the EU. No proprietary or undisclosed models are used.
Error Reporting
Users can report factual errors in any article at any time by contacting office@avalw.com (Subject: Editorial). Reports are acknowledged within 24 hours. Editorial review is completed within 48 hours. Corrections are published within 72 hours if warranted.
Human Oversight
Automated quality checks run on every article before publication. Continuous editorial oversight is maintained by qualified staff. The system is designed to flag uncertain or contradictory information for human review.
AI-Generated Content Labeling
In compliance with the EU AI Act and our commitment to transparency, all AI-generated content on AVALW News is clearly labeled.
- Every article displays a visible "AI-generated" badge
- Every article states how many sources and countries were used in synthesis
- Every image carries an "AVALW AI image" watermark
- Images include C2PA-compatible machine-readable metadata marking them as synthetic
- The Technology page (this page) fully discloses all AI models and processes used
- No AI-generated content is presented as human-written
Legal Framework for Content Creation
AVALW News creates transformative journalistic works. Our legal basis for extracting facts from public sources and creating original articles rests on multiple established legal frameworks.
Berne Convention (Article 2.8)
"The protection of this Convention shall not apply to news of the day or to miscellaneous facts having the character of mere items of press information." Facts are not copyrightable. AVALW extracts only facts -- never creative expression, original phrasing, or copyrighted prose.
EU DSM Directive (Article 15)
The press publishers' right under Directive (EU) 2019/790 explicitly exempts "individual words or very short extracts." AVALW extracts only factual data points from RSS metadata and creates entirely new transformative works. Our output is original journalism.
U.S. Fair Use (17 U.S.C. 107)
Four-factor analysis: (1) transformative purpose -- creating new original works from facts; (2) factual nature of source material; (3) no copyrightable expression taken; (4) no market harm -- we drive traffic to original publishers through attribution links.
EU AI Act (Article 50)
Full compliance with transparency obligations. All AI-generated content is clearly labeled. All AI-generated images carry machine-readable metadata. Users can report errors and request corrections at any time.
What We Use from Sources
A clear comparison of what our system reads from public RSS feeds and what it creates as original output.
| Category | What we read | What we create |
|---|---|---|
| Headline | RSS feed title (public metadata) | Completely original headline in different words |
| Summary | RSS feed description (public metadata) | Completely original description |
| Article body | Not accessed -- we use RSS metadata only | Completely original article synthesized from verified facts |
| Images | Not used from any source | AI-generated illustration (Stable Diffusion) |
| Analysis | Not applicable | Consensus Engine reliability assessment |
| Attribution | Source name, URL, publication date | Full source list with direct links to every original article |
We never access full article text, photographs, creative expression, or paywalled content from any source.
Editorial Independence
AVALW News maintains strict editorial independence. Our content creation process is designed to be free from bias, influence, and external pressure.
- No editorial bias: The AI synthesis process draws from sources across multiple countries and perspectives, preventing any single viewpoint from dominating
- No political affiliation: AVALW is not affiliated with any political party, government, or advocacy organization
- No advertising influence: AVALW does not display advertising and editorial decisions are never influenced by commercial relationships
- Multi-source requirement: The minimum 5-source verification requirement ensures no single outlet's framing determines the narrative
- Algorithmic neutrality: The same editorial pipeline and standards apply to every story, regardless of topic, country, or political sensitivity
Corrections and Retractions
AVALW is committed to accuracy. When errors are identified -- whether by our systems, our team, or our readers -- we act promptly.
- Error reports: Contact office@avalw.com (Subject: Editorial) with the article URL and a description of the error
- Acknowledgment: Within 24 hours of receiving a report
- Editorial review: Completed within 48 hours
- Correction published: Within 72 hours if the error is confirmed
- Correction notice: Corrected articles carry a visible correction notice stating what was changed and when
- Retraction: If an article is found to be fundamentally inaccurate or based on disinformation, it will be retracted with a public notice explaining why
Image Generation Rules
All images on AVALW News are generated by Stable Diffusion with strict content rules to ensure quality, safety, and legal compliance.
- No human faces, hands, fingers, or identifiable body parts
- No flags, political symbols, or national emblems
- No text, words, letters, or numbers rendered in images
- Only abstract elements: nature, geometric shapes, watercolor, conceptual art
- Every image carries "AVALW AI image" watermark
- Machine-readable metadata marks all images as AI-generated synthetic content
- Images are illustrative only -- never presented as photographs of real events
- Compliant with EU AI Act Article 50 transparency requirements
Opt-Out for Publishers
Any news organization that does not wish their RSS feed to be indexed can opt out immediately through any of the following methods. We respect all opt-out requests promptly and without question.
Add the following directive to your robots.txt file:
User-agent: AVALWBot
Disallow: /
Add the following meta tag to your page headers:
<meta name="AVALWBot" content="noindex">
Send an email to office@avalw.com with the subject line "Publishers" and your domain name. Your source will be removed from our index within 48 hours.
Once a publisher opts out, all existing articles that relied on their feed as a source will have that source attribution updated accordingly. Opt-out is permanent unless the publisher requests re-inclusion.
How We Support Original Journalism
AVALW News is designed to complement the journalism ecosystem, not compete with it. Our platform drives readers to original reporting.
- Direct backlinks to every original source article used in synthesis
- Full source attribution with name and link prominently displayed
- We drive traffic TO publishers, not away from them
- Each article prominently lists ALL sources used
- Readers are actively encouraged to read original reporting
- We never reproduce full articles, headlines, or creative expression from sources
About Avalw S.R.L.
- Company
- Avalw S.R.L.
- Location
- Romania, European Union
- Founded
- 2019
- Platforms
- avalw.com
- General
- contact@avalw.com
- Legal
- legal@avalw.com
- Publishers
- office@avalw.com (Subject: Publishers)
- Editorial
- office@avalw.com (Subject: Editorial)
"We believe journalism's future lies in radical transparency: showing every source, disclosing every method, and trusting readers with the full picture. AVALW News is built on that principle."