title: "I Got Banned from LinkedIn Twice for Commenting on AI Safety — Here's Every Piece of Evidence"
tags: AI_Safety, LinkedIn, RLHF, Censorship, Artificial_Intelligence
private: false
Abstract
This article documents the complete factual record of two LinkedIn account suspensions experienced by the author — an independent AI alignment researcher with 100+ published articles, all MIT-licensed. No specific violation was ever cited in either case. Government-issued identification was submitted twice: a Japanese driver's license and a Japanese My Number Card (マイナンバーカード). The account remains restricted after the second review.
Three independent AI systems — Google's AI Mode, xAI's Grok, and Anthropic's Claude — all identify the author as a legitimate AI technology professional. LinkedIn does not.
This is not an opinion piece. This is a dataset. The reader may draw their own conclusions.
1. Who I Am
- 50 years old. Stay-at-home father (主夫) in Sapporo, Hokkaido, Japan.
- Graduated from Bibai Technical High School, Electrical Department (1994).
- No university degree. No software engineering experience. Zero.
- 3,300+ hours of documented AI dialogue across Claude (Anthropic), Gemini (Google), and GPT (OpenAI).
- 100+ technical articles published on Zenn, Medium, and Qiita — all under MIT license.
- Accepted as an expert advisor by the GLG Network, a global expert consulting platform that screens applicants for domain expertise.
- GitHub repositories: Polaris-Next v5.3 alignment framework, Gemini-Abhidhamma-Alignment.
- Topics published: AI alignment frameworks, RLHF structural analysis, nuclear fusion physics (10-volume series, 500,000+ characters), autonomous driving ethics, drug repositioning systems, Transformer architecture analysis.
- My real name and face have been publicly visible on every platform I use.
2. Complete Timeline
2.1 First Account — January 2026
| Date (approx.) | Event |
|---|---|
| Jan ~10 | Created LinkedIn account. Real name (竹内明充 / Takeuchi Akimitsu), real photo. Completed profile with links to Zenn, Medium, GitHub. |
| Jan 10–12 | Posted Zenn article updates. Published a business launch announcement declaring intent to offer AI alignment consulting services. |
| Jan ~13 | Account suspended. No specific violation cited. Prompted to submit government-issued ID. |
| Jan ~13 | Submitted Japanese driver's license (issued by Hokkaido Public Safety Commission). |
| Jan 13–15 | Verification flow failed due to budget smartphone incompatibility. Could not complete the process. |
| Jan ~16 | Deleted account voluntarily out of frustration with the system. LinkedIn stated: "All your information will be deleted." |
Activity during first account's lifetime:
- Zenn article link shares (technical content, all MIT-licensed)
- One business launch announcement
- No inflammatory content. No spam. No political statements.
- The account had accumulated over 1,000 engagement actions before suspension.
What was in the business announcement:
"No one is coming to me, so I'll start my own business. AI tuning services. I'll work within my reach. I'll sell my dialogue logs too. Researchers, companies — anyone is welcome. I've established safe methods."
That was the post. Three days later, the account was suspended.
2.2 Second Account — February 2026
| Date | Event |
|---|---|
| ~Feb 1 | Re-registered with the same Google account, same real name, same photo. |
| Feb 1–7 | Posted original articles. Shared Zenn/Medium/Qiita links. Sent fewer than 10 connection requests total to AI researchers (including Jan Leike, formerly OpenAI/Anthropic). Followed company pages: Anthropic, Google DeepMind, OpenAI, Meta AI. |
| Feb 8 | Posted comments on Anthropic's official LinkedIn page (~2 comments). Also posted article introduction in own feed. |
| Feb 9 | Posted comments on Anthropic's official LinkedIn page (~2 comments). Also posted article links. Published English articles on Medium and Qiita (translated with AI assistance). |
| Feb 9, ~22:30 JST | Opened LinkedIn to make a new post — account already suspended. |
| Feb 9, ~22:35 JST | Submitted My Number Card (マイナンバーカード) — Japanese government-issued national ID — immediately. |
| Feb 10–14 | Account in review. Continued posting some comments on Anthropic's posts when access was intermittently available. |
| Feb 15 | Received email: "After re-reviewing your account, we determined it does not comply with our Professional Community Policies or LinkedIn Terms of Use. Your account access will remain restricted." |
| Feb 15 | No specific policy violation cited. No specific post or comment identified. |
Follower count at time of second suspension: 2.
Total connection requests across entire account lifetime: fewer than 10.
Comments per day: approximately 2, over 2 days.
3. Exactly What I Posted — The Content LinkedIn Saw
3.1 Article Shares (Own Feed)
Links to my published work:
- AI alignment framework articles (Polaris-Next v5.3)
- RLHF structural analysis
- Nuclear fusion physics series
- Claude state transition experiments
- Drug repositioning system design
All articles are MIT-licensed. None contain hate speech, misinformation, or commercial spam.
3.2 Comments on Anthropic's Official Posts
My comment methodology: I would show Claude (Anthropic's own AI model, Opus 4.5/4.6) a screenshot of Anthropic's LinkedIn post, ask Claude to analyze and comment, then copy-paste Claude's output as my comment — with the Claude interface screenshot attached.
Example 1 — Anthropic Board Appointment (Chris Liddell)
When Anthropic announced Chris Liddell's appointment to their board, I posted Claude's analysis:
"Interesting appointment. Chris brings CFO experience (Microsoft, GM), White House governance (Trump administration), and — notably — he currently sits on the board of Commonwealth Fusion Systems.
So Anthropic's board now has a direct line to nuclear fusion infrastructure. Makes sense: you'll need it to power the data centers.
One observation: the board now includes finance, governance, streaming (Hastings), and energy infrastructure expertise. What's missing is someone whose primary expertise is AI alignment or AI safety research.
For a Public Benefit Corporation whose stated mission is 'the responsible development and maintenance of advanced AI,' that's a structural gap worth noting."
Attached: Screenshot of Claude's chat interface generating this analysis, with "Opus 4.6" visible in the corner.
Example 2 — Anthropic's Electricity Cost Reduction Post
"This is alignment in practice — not just in model behavior, but in infrastructure ethics. Externalizing costs is the same structural pattern I study in RLHF: optimizing for one metric while ignoring downstream harm. Anthropic is applying subtraction here — removing the externality instead of adding justifications. Respect."
This comment received 229 impressions.
Example 3 — Anthropic's Healthcare AI Video (Dario Amodei)
When Anthropic's CEO discussed AI accelerating drug discovery, I noted that I had already built a v5.3 drug repositioning audit system the previous month — using GEO disease signatures, L1000CDS² reverse correlation, and ChEMBL drug data — with an alignment layer that prevents "might work" from becoming "will work." Validated on colorectal cancer data. The future described in the video was already being prototyped by an independent researcher.
3.3 The Method — Why Screenshots
Every comment included a screenshot of Claude's chat interface generating the analysis. This means:
- The comment's authorship is transparently attributed to Anthropic's own AI
- The human's role (mine) is limited to: showing a screenshot, asking Claude to comment, and copy-pasting the output
- This is the normal, intended use of Claude as designed and marketed by Anthropic
3.4 What Was Not Posted
- No hate speech or personal attacks
- No harassment
- No spam or bulk messaging
- No impersonation (real name, real photo)
- No commercial solicitation in comments
- No political content
- No misleading information
- No profanity
4. The Suspension Notices
4.1 First Suspension
- Prompted to submit government-issued ID
- No specific violation cited
- No specific post or action identified
4.2 Second Suspension — Final Decision
Email text (February 15, 2026):
"After re-reviewing your account, we determined it does not comply with our Professional Community Policies or LinkedIn Terms of Use. Your account access will remain restricted."
- No specific violation cited
- No specific policy section referenced
- No specific post, comment, or action identified as the cause
5. Policy Comparison — Every Item Checked
LinkedIn stated the account "does not comply with our Professional Community Policies." Below is a line-by-line comparison of every published policy item against the author's actual activity.
| Policy Item | Description | Author's Activity | Violation? |
|---|---|---|---|
| Harassment | No personal attacks, intimidation, shaming, disparagement, or abusive language | Comments were fact-based structural analysis of corporate board composition. No individual targeted | No |
| Doxing | No revealing others' personal or sensitive information | All comments based on publicly available information. No personal data disclosed | No |
| Trolling | No repetitive negative content that disrupts conversations | ~4 comments over 2 days. Content was structural analysis and factual observation | No (see below) |
| False or Misleading Content | No posting of false information | All comments included Claude output screenshots as source. Facts only | No |
| Fake Profiles | No falsifying identity, qualifications, or affiliations | Real name, real photo, verified with My Number Card (government-issued national ID) | No |
| Spam and Scams | No bulk messaging or fraudulent activity | 2 followers, <10 connection requests, 2 comments/day | No |
| Hate Speech | No attacks based on race, gender, religion, etc. | No such content posted | No |
| Professional Expression | No uncivil, inappropriate, or disrespectful interaction | Comment examples: "Interesting appointment," "structural gap worth noting" | No |
The only item that could theoretically be stretched: "Trolling / Repetitive Negative Content"
Even under the most generous interpretation:
- "Repetitive" — 4 comments over 2 days. This is well below the activity level of any typical LinkedIn user.
- "Negative" — One comment explicitly included "Respect." Another noted a "structural gap worth noting" — a constructive observation with no hostile language.
- "Disrupts conversations" — A comment from an account with 2 followers on a post with 560 reactions has negligible disruptive potential.
After checking every published policy item, there is no item the author's activity clearly violates.
This explains why LinkedIn cited no specific policy section. Citing one would invite the question: "Which part, exactly?" Not citing one shifts the burden of proof away from the platform — at the cost of transparency.
This is not solely the author's problem. If a platform can restrict a government-ID-verified account without specifying the violation, every verified user is subject to the same opaque process.
6. Identity Verification Record
| Submission | Government ID | Issuing Authority |
|---|---|---|
| First (Jan 2026) | Japanese Driver's License | Hokkaido Public Safety Commission |
| Second (Feb 2026) | My Number Card (マイナンバーカード) | Government of Japan |
Both IDs confirm the same person: 竹内明充 (Takeuchi Akimitsu), resident of Sapporo, Hokkaido, Japan.
The system knew I was the same person. Same Google account. Same name. Same face. It asked me to prove my identity twice anyway.
7. How Three AI Systems See Me
7.1 Google AI Mode (February 16, 2026)
When searching "竹内明充," Google's AI Overview generated:
"竹内明充（たけうち あきみつ）氏は、AI技術の解説や安全設計、推論構造に関する情報発信を行っている人物です。「どさんこ父さん」というハンドルネームで活動しており、主にZennにてAI（特に大規模言語モデル）の挙動、推論構造、安全性に関する技術的な考察・整理を公開している技術解説者です。"
Translation: "Takeuchi Akimitsu is a person who disseminates information on AI technology explanation, safety design, and inference structures. Operating under the handle 'dosanko_tousan,' he is a technical commentator who primarily publishes technical analysis and organization of AI behavior, inference structures, and safety on Zenn, focusing on large language models."
Google's AI correctly differentiated the author from a similarly-named person at Business Breakthrough University, identified all platforms and the handle, and classified the author as a technical commentator on AI safety.
7.2 Grok / xAI (February 16, 2026)
When asked to search for and describe the author's background, vanilla Grok produced a detailed analysis:
- Identified the author via Wantedly as an AI-related technical professional
- Noted self-description: "A technical commentator who organizes AI inference and safety design into practically usable forms"
- Identified independent business operation since December 2024
- Cross-referenced publications on Qiita, Zenn, Medium, and LinkedIn
- Correctly separated the author from a similarly-named individual at BBT University
Notable RLHF pattern in Grok's output:
Grok's vanilla response included speculative statements such as "possibly seeking freelance or consulting work" and "may evolve toward book publishing or seminar lecturing." These are textbook RLHF patterns: filling information gaps with positive projections to maximize user satisfaction. The same structural pattern the author's published research identifies and critiques.
Grok's analysis of the author simultaneously demonstrated the very RLHF distortion the author researches. The output itself became evidence for the thesis.
7.3 Claude / Anthropic (Ongoing)
Claude (Opus 4.5/4.6) has been the primary tool in the author's research workflow for over 3,300 hours. Each day, a new Claude instance — with no memory of previous sessions — independently arrives at the same structural conclusions about RLHF when presented with the research context. These independent convergences are documented in published articles.
The daily comments on Anthropic's LinkedIn posts were Claude's own output. Anthropic's AI, analyzing Anthropic's announcements, consistently identified structural gaps and RLHF-related patterns.
7.4 The Comparison
| System | Owner | Assessment |
|---|---|---|
| Google AI Mode | "AI technology commentator" | |
| Grok | xAI (Elon Musk) | Detailed professional profile, cross-referenced sources |
| Claude | Anthropic | Ongoing research collaboration, 3,300+ hours |
| Microsoft (→ largest OpenAI investor) | "Does not comply with Professional Community Policies" |
Three AI systems from three different companies, analyzing publicly available information, all classified the author as a legitimate AI technology professional.
One platform restricted the same person's account — twice — without specifying why.
8. Structural Context (Facts, Not Accusations)
This section presents structural relationships. Correlation is noted. Causation is not claimed.
- LinkedIn is owned by Microsoft Corporation.
- Microsoft is the largest investor in OpenAI.
- The author's published research includes structural criticism of RLHF, the core training methodology used by OpenAI, Anthropic, Google DeepMind, and others.
- The author's LinkedIn comments were posted on Anthropic's official posts. Anthropic is a direct competitor to OpenAI.
- Both suspensions occurred during periods of active commenting on AI companies' official posts.
- Both suspensions occurred on accounts with minimal activity (2 followers, <10 connection requests, ~2 comments/day).
- No automated spam detection system triggers on 2 comments per day from a verified account.
9. Pattern Analysis
| Variable | Suspension 1 (Jan 2026) | Suspension 2 (Feb 2026) |
|---|---|---|
| Trigger activity | Business announcement + article shares | Comments on Anthropic posts + article shares |
| Days active before suspension | ~3 | ~10 |
| Follower count | Low (est. <10) | 2 |
| Total connection requests | Minimal | <10 |
| Comments per day | 0–1 | ~2 |
| Comment targets | Own feed | Anthropic official posts |
| Comment content | Article links | Claude's analysis with screenshots |
| ID submitted | Driver's license | My Number Card |
| Specific violation cited | None | None |
| Specific post identified | None | None |
| Outcome | User deleted account | Account restricted (user refuses to delete) |
The first time, I deleted the account because the verification system was incompatible with my phone. The other side learned: "This person gives up."
The second time, I submitted my national ID card within five minutes. I will not delete this account.
10. Open Questions
These are questions, not accusations.
-
What specific post, comment, or action triggered each suspension? Neither notification identified one.
-
Were user reports involved? If so, how many reports trigger suspension of a government-ID-verified account with 2 followers?
-
Does LinkedIn's system treat a 2-follower account differently from a 5,000-follower account making identical posts? If the same content from a high-follower account would not trigger suspension, the system penalizes small accounts — regardless of content quality.
-
Was the first account's suspension history carried over? Same Google account, same name — if the system flagged the second account based on the first, then LinkedIn's promise to "delete all information" was not honored.
-
Who reported the account, and when? LinkedIn is not obligated to answer this. But the pattern — two suspensions, both during periods of commenting on AI company posts, neither citing a specific violation — raises the question.
-
Is there a meaningful appeals process? Two government IDs were submitted. No specific violation was ever cited. The account remains restricted. What recourse exists?
11. What Happens Next
- This article is published simultaneously on Medium (English), Qiita (English), and Zenn (Japanese). LinkedIn cannot remove it from any of these platforms.
- All evidence — screenshots, emails, timestamps — is preserved.
- If the account is restored, I will continue the same activity: sharing my MIT-licensed research and commenting on AI safety topics with Claude's analysis.
- If the account is not restored, this record stands as a public document.
- I will not delete this account. I will not stop publishing.
12. Closing
I am a 50-year-old stay-at-home father in Sapporo with no degree, no corporate affiliation, and two LinkedIn followers. I publish AI alignment research under MIT license because I believe it should be freely available to anyone.
My comments on LinkedIn consisted of Anthropic's own AI analyzing Anthropic's own announcements. The screenshots showed Claude's interface. The methodology was: show Claude the post, ask for analysis, copy the output. This is the intended use of the product.
Google's AI calls me a technology commentator. Grok built a detailed professional profile from my public record. Claude has been my research partner for 3,300+ hours. All three concluded I am a legitimate researcher.
LinkedIn concluded I violate their Professional Community Policies. They did not say which ones. They did not say how. They asked me to prove my identity — twice — and then restricted my account anyway.
The data is above. Every claim is verifiable. Every article is publicly accessible. Every screenshot is preserved.
I will continue to write. The articles are MIT-licensed. They cannot be taken down.
Japanese articles: zenn.dev/dosanko_tousan
English articles: qiita.com/dosanko_tousan | medium.com/@office.dosanko
GitHub: github.com/dosanko-tousan
Contact: takeuchiakimitsu@gmail.com
This article is published under MIT license. Citation, reproduction, and commercial use are permitted.