A model that confidently misreads a conversation could trigger a real-world consequence: a false welfare check, a flag that follows someone, with enormous weight on a user’s life, legally, socially, and politically.
The cases feel straightforward. This question isn’t. 3/3
What does it look like for AI to be a mandated reporter, legally obligated, like a therapist or teacher, to flag certain things a user discloses?
Florida’s AG suing OpenAI, along with a few other recent chatbot cases, has me thinking past the immediate legal questions. 1/3
Florida’s AG suing OpenAI, along with a few other recent chatbot cases, has me thinking past the immediate legal questions. We tend to assume a reporting requirement makes these tools safer. But we rarely account for hallucination. 2/3
Fable 5 issue as a show of strength and activity around these concerns. Especially since almost no technical staff or leadership had a part in the export control gauntlet being thrown down. (3/3)
IS FABLE 5 THE SACRIFICIAL LAMB FOR TRUMP’S MIDTERM FEARS?
I’ve been waiting to comment on the Fable 5 back and forth to see what information continues to come out and the more I hear, the more I feel like this is more about the midterms than it is “national security”. (1/3)
With poll after poll coming out about how around 70% of average Americans feel that the government isn’t doing enough or can’t be trusted to regulate or govern AI and an approval rating of less than 40%, I wouldn’t put it past this administration to be using this (2/3)
SCOOP: The White House invited online safety groups for another meeting on Sen. Blackburn's forthcoming AI preemption/online safety package, per 2 ppl familiar and an invite obtained by POLITICO. Meeting set for 3pm today.
w/ @Owen_Dahlkamp
Does your org work on AI evals?
@NIST is taking letters of interest to collaborate on AI standards.
Its AI Consortium will have 6 task groups, including:
- AI Testing, Evaluation, Verification & Validation
- Annotation for AI Risks & Validity
- Chemical & Biological Security
local backlash against data centers.
Seems like the administration will do just about anything other than demand comprehensive legislation from Congress around artificial intelligence and its impact (2/2)
Pam Bondi being added to the President’s Council of Advisors on Science and Technology in a liaison capacity between government agencies and the council may signal a more active involvement in the preemptive efforts against state AI legislation, especially relating to the (1/2)
Illinois just passed SB 315 -- with unanimous House support and a 52-5 Senate vote -- making it the first state to require independent third-party audits of frontier AI models. It also joins California and New York in setting AI safety standards.
This comes days after industry pressure led the White House to shelve its own AI safety testing framework. Another reminder that state legislators are responding to the bipartisan demand for AI safeguards, while the American people wait for Washington to act.
https://t.co/Szu7LBFhGu
FULL timeline of how the AI exec. order came together:
- Since late 2025, there’s been a group in the admin working on a framework for AI and cybersecurity
- This February pre-Mythos, a draft EO was circulated across agencies
- In late Feb/March, admin staffers began briefing Wiles, Bessent and JD Vance based on what they heard from informal networks about Anthopic’s model
- Mythos’ public arrival in April started a number of senior officials including Wiles and Bessent who are concerned abt the risks posted to critical infra.
- In late April, ONCD began working with OpenAI, Google, and Anthropic, making them sign NDAs ahead of one-on-one and joint meetings
- The three companies and their government affairs teams provided feedback on the AI framework and the language in the EO
- Aside from industry, the new EO was closely held and did not go thru interagency process
- There were two table reads on May 20, where admin brought AI execs to EEOB to review hard copy
- Everyone appeared in agreement, signing ceremony was planned
- Then David Sacks came in on morning of May 21, called President Trump and tanked it
w/ @diana_nerozzi
https://t.co/Bydl6CcFGG
This week in AI policy: 1) Trump's AI executive order collapsed hours before signing. 2) SSCI marked up FY27 IAA with new IC AI oversight, 14-3. 3) FTC closed $930K AI-washing case vs Cox Media Group. 4) Illinois, Ohio, SC, MN, CA all moved state AI bills.
White House may be getting cold feet on pre-testing frontier models. If the big labs are already doing it voluntarily, then what could the real issue be? Real accountability?
https://t.co/cK6zpfuHdm
NEW: The Senate has unanimously passed the Stop Stealing Our Chips Act.
This bill is an important step towards strengthening export control enforcement in face of growing smuggling operations. America must retain its compute advantage.
Our statement: https://t.co/Uj0L6S38mk