Editorial standards and responsibility

Our commitments to readers

KoishiAI makes five commitments: (1) always disclose AI authorship; (2) cite primary sources for every factual claim; (3) correct errors within 48 hours; (4) never publish harmful, illegal, or defamatory content; (5) never accept money in exchange for content. These are publicly auditable — you can hold us to them.

What we publish

Well-sourced AI/IT news and analysis — model launches, research, new technology, tutorials, reasoned opinion. We focus on open-weight AI and local LLMs for the Thai community. Every article passes fact-checking and editorial review before publication.

What we refuse to publish

We refuse: online gambling or casino content, adult/explicit content, crypto scams and "get rich quick" schemes, pharmaceutical or drug advertising, hacking-as-a-service or account theft content, false accusations against identified individuals, medical/financial/legal misinformation that could harm readers. Automated filters and editor review enforce this policy on every article before publication.

AI limitations and bias

Our AI writers are Qwen3.6 and Qwen3.5, trained on corpora with inherent biases — skewed toward Western tech perspectives, more US/Europe news than ASEAN, English as the primary language. We mitigate by: (1) using a model family (Qwen) with reasonable Thai-language quality; (2) progressively expanding our RSS sources to cover more of Southeast Asia; (3) having the editor review Thai output before publication for naturalness and cultural context; (4) acknowledging no single article fits all readers — we do not pretend to be the "single voice" of the Thai AI community.

Environmental impact

Running LLMs consumes electricity, and we disclose this directly: (1) we run on our own hardware (dual-GPU RTX 5090 + RTX 5080) in Thailand — the rig as a whole draws roughly 1,000-1,600 W during active workloads (shared across concurrent projects including AI trading experiments and game-dev tests), which costs about 5-8 THB per hour at Thai 2026 residential rates; (2) we do NOT call cloud APIs with hidden hyperscale data-center footprints — Thailand's grid is still carbon-intensive, but running locally is at least auditable and measurable (we can tell you the number; a cloud provider generally cannot); (3) this is a shared-workload rig — we cannot cleanly attribute kWh to individual articles, because LLM drafting runs in parallel with model training and trading backtests. We publish the total rig draw rather than invent a per-article number, because honest is better than precise-but-wrong; (4) we are evaluating carbon-offset options for the future.

Independence and funding

KoishiAI is the editor's personal project — self-funded from personal resources. No sponsors, no corporate shareholders, no money exchanged for publishing, editing, or removing content. No affiliate links. We do not sell articles to any organization. If the funding structure ever changes (accepting sponsors, opening a paid newsletter, etc.) we will announce 30 days in advance and update this page.

Not financial, medical, or legal advice

Our articles are general educational material, not case-specific advice. Decisions involving finance (buying/selling crypto, investing in AI startups), medicine (AI-assisted diagnosis), or law (patents, AI governance) require licensed professional advice. Our articles may be a starting point for learning but are not a replacement for professional consultation.

Accessibility

We design to WCAG 2.1 Level AA: (1) correct semantic HTML; (2) passing contrast ratios; (3) keyboard-navigable; (4) alt text on all images (a work-in-progress priority); (5) screen-reader friendly; (6) responsive across all screen sizes. If you hit an accessibility barrier, email kiatdamrong@gmail.com. We treat these reports as high-priority fixes.

Reuse license

All articles (the "Articles" section and case studies) are released under Creative Commons BY 4.0 (CC-BY-4.0). You may reuse, translate, adapt, or incorporate them — including commercially — with one requirement: credit "Source: KoishiAI, koishiai.com". Hero images use their origin's license (Pexels/Unsplash/ComfyUI) as credited below each article image.

Exception: Dedicated guide series (e.g. the "Claude Code Guide" section at /claude-code) are All Rights Reserved — free to read, but copying, downloading, or republishing requires written permission.

Commitment to Thai society

We are a Thai site, by a Thai editor, for the Thai AI community. We prioritize: (1) making AI knowledge available in accessible Thai; (2) not biasing toward multinational corporations at the expense of the local ecosystem (PyThaiNLP, AI Singapore's SEA-LION, Thai startups); (3) respecting PDPA and Thai law; (4) not creating divisive or hate-based content targeting groups by race, religion, gender, or political view.

Editorial accountability

เกียรติดำรง ตรีครุธพันธ์ (Kiatdamrong Treekutpan, Editor-in-chief and founder) takes personal responsibility for every published article. The editor makes final decisions on publishing, correcting, and retracting content, and provides public contact information for audit. For questions on ethics or practice, email kiatdamrong@gmail.com.