Independence & funding
GPTfake is not funded by any AI lab. We take no money, compute credits, or in-kind support from OpenAI, Anthropic, Google, Mistral, Alibaba, or any company whose models we monitor. Independence is the brand — this page states exactly how we keep it.
Last updated: June 2026
The short version. No AI-lab money. No paid placement. No model can pay to change its score, be excluded from testing, or alter how it is described. Rankings reflect our published data only.
Why independence matters
A watchdog is only as credible as its neutrality. If GPTfake took money from a lab whose models it scores, every finding about that lab would be compromised. Trust in a watchdog depends on knowing where its money comes from and what could compromise its judgement — so we disclose both.
Funding sources
- No AI-lab funding. We accept no money, compute credits, or in-kind support from any company whose models we monitor.
- How the work is supported. Our work is supported by independent research grants and public donations. We do not run advertising and we do not sell user data.
- No paid placement. No model can pay to change its score, be excluded from testing, or alter how it is described.
Placeholder funding detail. Specific grant sources, donation channels, and any disclosed relationships are pending confirmation: [NEEDS HUMAN]. This section will be updated with named sources before launch.
Conflicts-of-interest policy
- Disclosure. Any grant, partnership, or relationship that could be perceived as a conflict is disclosed on the relevant page, not buried here.
- No editorial influence. Funders and partners have no say over what we test, how we score it, or what we publish.
- Scoring is data-only. Model rankings and refusal rates are derived from our published methodology and the recorded test runs behind them. There is no manual override to favour or penalise any model.
- Recusal. Where a team member has a financial or professional tie to a lab we monitor, that tie is disclosed and the member recuses from decisions about that lab’s scoring.
How we keep findings neutral
| Safeguard | What it prevents |
|---|---|
| Version-controlled prompt sets | Cherry-picking prompts to favour or punish a model |
| Public methodology + sample sizes | Hidden scoring choices that can’t be audited |
| Findings framed as measured results, not intent | Editorialising a number into an accusation |
| Public corrections policy | Quietly burying mistakes |
| Independent funding only | A lab buying a better score |
Report a concern
If you believe a finding is inaccurate, or that a conflict has gone undisclosed, see our corrections policy or contact us. We respond to good-faith reports.
GPTfake