The AI Whistleblower Initiative

The AI Whistleblower Initiative (AIWI) is a Berlin-based independent nonprofit founded in 2024 by Karl Koch to support concerned employees at frontier AI companies who wish to raise safety and ethical concerns. Launched initially as OAISIS before rebranding in July 2025, AIWI operates as a specialized intermediary: it sits between AI insiders who hold safety-critical knowledge and the legal, technical, and financial infrastructure those insiders need to act on it safely. The organization combines direct insider support — anonymous consultation, legal-defense referrals, operational-security guidance — with public-accountability campaigns that pressure frontier AI companies to establish and publish transparent internal whistleblowing channels.

Founding and research base

Karl Koch began the research underpinning AIWI in early 2024, consulting more than 100 AI governance researchers and insiders before launching the organization's first programs. His background spans management consulting, technology entrepreneurship (a SaaS business acquired in 2023), and AI governance research — he had been a volunteer researcher at the Future of Humanity Institute since 2016. The consultation phase surfaced a structural gap: AI insiders held safety-relevant information and faced a specific stack of deterrents — NDAs, non-disparagement clauses tied to equity vesting, industry blacklisting — that generic whistleblower support organizations were not equipped to navigate. AIWI was purpose-built to fill that gap.

Third Opinion — anonymous pre-whistleblowing consultation

AIWI's flagship program, Third Opinion, launched in late 2024 as a structured pre-whistleblowing service. AI insiders can submit questions about their safety or ethical concern anonymously through a Tor-based, open-source tool — without disclosing the employer's name or any confidential information. AIWI assembles a panel of independent technical and governance experts customized to the insider's specific question and facilitates an anonymous Q&A. The tool is pen-tested and its security reports are published for public scrutiny. If the consultation surfaces a concern that warrants further action, AIWI escalates the connection to pro bono legal counsel with client-privilege protection.

The program addresses a finding from AIWI's insider survey: approximately 50% of AI employees lack confidence in their own ability to assess whether their concern is serious enough to justify disclosure. Third Opinion explicitly lowers the personal judgment threshold — providing expert grounding before an insider commits to a legally and professionally costly course of action.

Supporting services and partner network

Beyond Third Opinion, AIWI offers support spanning the disclosure lifecycle. It refers insiders to partner organizations — including the Government Accountability Project, the Signals Network, and WHISPeR — for legally privileged representation specialized in whistleblower rights, risks, and options. It distributes operational-security guidance including digital-privacy guides and hardened-device recommendations for insiders communicating about sensitive concerns. And it assists with locating financial support for legal costs.

AIWI's insider survey documented the landscape this infrastructure addresses: more than 90% of surveyed AI employees could not identify any whistleblower support organization; 55% were unaware of internal reporting channels at their own employer despite policies existing; and 100% expressed low confidence that regulators understand AI-specific concerns well enough to act on reports.

Publish Your Policies campaign

On July 30, 2025 — National Whistleblower Day — AIWI launched its first major public-accountability campaign, calling on frontier AI companies to publish their internal whistleblowing policies. At the campaign's launch, OpenAI was the sole frontier AI company that had done so; Anthropic, Google DeepMind, Meta AI, xAI, and Mistral had not.

The campaign specified two tiers of transparency: Level 1 (minimum) — full public disclosure of whistleblowing policy scope, reporting channels, investigation procedures, and protection measures; Level 2 (aspirational) — effectiveness metrics including report volumes, case outcomes, retaliation complaints, and independent audit results. More than 30 civil-society organizations signed on, including Transparency International, the National Whistleblower Center, the Government Accountability Project, the Future of Life Institute, the Center for AI Policy, and Blueprint for Free Speech. Academic endorsers included Stuart Russell (UC Berkeley), Lawrence Lessig (Harvard Law School), Roman Yampolskiy (University of Louisville), Daniel Kokotajlo, and Nathan Labenz. At least one unnamed frontier AI company responded to AIWI's outreach with a cease-and-desist letter.

Policy advocacy

AIWI has positioned itself as a subject-matter expert in AI-specific whistleblower legislation and regulation. Koch co-authored commentary on whistleblower-protection provisions in California's AI safety bill SB 53 and produced a resource on whistleblowing under the EU AI Act. He contributed subject-matter expertise to the EU AI Office during the development of the world's first AI-specialized whistleblowing channel — activated under Article 87 of the EU AI Act in August 2026, which formally applies the EU Whistleblower Directive (2019/1937) to reports of AI Act infringements. In the United States, AIWI supports the pending AI Whistleblower Protection Act, which would extend statutory protection to AI-industry disclosures at the federal level.