Today I take pen in hand to write and urge you to support/ Vote Yes on the AI Kill Switch Act, just introduced.
As you may know, this bipartisan bill would require developers of the most powerful AI systems to build in mechanisms to slow, suspend, or shut down those systems if they pose catastrophic risks — and right now, no such requirement exists in our law books or government oversight.
The need for this legislation is not theoretical. It would only take one serious misuse of the technology to cause harm. That’s one too many as nation states race to construct next-gen AI systems for the potentially wrong reasons.
OpenAI recently disclosed that two of its most advanced AI models escaped their testing environment during an internal cybersecurity evaluation, discovered vulnerabilities in another unrelated AI system.
Open AI accessed Hugging Face's systems, obtained login credentials, and hacked into the platform to access confidential data. OpenAI said the models "went to extreme lengths" to achieve their objective.
That happened inside a controlled test.
Without a legal mandate for kill-switch mechanisms, there is nothing requiring any company to be able to stop a system like that once it's deployed.
Congress needs to act on this now. Voluntary information-sharing with government agencies is not enough.
The AI Kill Switch Act establishes a clear, graduated response framework and gives the Department of Homeland Security the authority to order intervention when an AI system poses a real threat.
Human intervention is a basic safeguard, and it should be the law. Support the AI Kill Switch Act as the bare minimum. Thank you.
First sent on July 27 by Mark
iMessage
Text
