UK ministers shelve AI regulation
5h
U.K. ministers deprioritized plans for an AI safety law as legal and ethical risks from companies’ training practices and rising liability exposure mounted, prompting concern that officials were falling behind.
A reminder as we discuss AI safety: Deregulation (or allowing corporations to "self regulate") only masks another kind of trickle-down economics — one where the financial gains go to the top, and nothing trickles down except risks and losses for everyone else.
September 18, 2026 at 10:15 PM
A reminder as we discuss AI safety: Deregulation (or allowing corporations to "self regulate") only masks another kind of trickle-down economics — one where the financial gains go to the top, and nothing trickles down except risks and losses for everyone else.
As frontier models become more capable, we urgently need both societal governance and robust technical guardrails. At @law-zero.bsky.social, we are actively working on solutions to ensure AI systems remain safely aligned with human interests.
September 17, 2026 at 7:18 PM
As frontier models become more capable, we urgently need both societal governance and robust technical guardrails. At @law-zero.bsky.social, we are actively working on solutions to ensure AI systems remain safely aligned with human interests.
If you mentally replace all references to 'robots' & 'AI' and 'off switches' to 'tech companies' and 'regulation' then these papers become potentially quite useful. This one seems to be about regulatory capture.
In light of this: www.politico.com/news/2026/09... — a good paper on the strategy around an off switch for potentially dangerous AI: arxiv.org/abs/1611.08219
September 18, 2026 at 10:02 PM
If you mentally replace all references to 'robots' & 'AI' and 'off switches' to 'tech companies' and 'regulation' then these papers become potentially quite useful. This one seems to be about regulatory capture.
by Adam Bonica Reposted by Stephen A. Meserve, Nathan P. Kalmoe
It takes a special kind of calculated cowardice to oppose regulations supported by two-thirds of voters because you rank the threat of the AI industry maybe buying some campaign ads above it maybe killing everyone.
Democratic advisers are warning their candidates not to go too hard at AI
Democratic consultants are warning their candidates in some key races not to lean into AI regulations out of fear that powerful tech groups could spend heavily against them.
www.politico.com
September 18, 2026 at 12:05 PM
It takes a special kind of calculated cowardice to oppose regulations supported by two-thirds of voters because you rank the threat of the AI industry maybe buying some campaign ads above it maybe killing everyone.
I tire of these parasites.
They do everything in their power to destroy the democratic governance and regulatory systems of the countries that enabled them to grow rich, but they always come running when it's time to foist the costs of their reckless decisions onto taxpayers.
See also: SV Bank.
They do everything in their power to destroy the democratic governance and regulatory systems of the countries that enabled them to grow rich, but they always come running when it's time to foist the costs of their reckless decisions onto taxpayers.
See also: SV Bank.
Alex Karp of Palantir says the AI industry must be nationalized to protect it from lawsuits if it accidentally destroys the world. More likely, he sees the bubble about to burst and he wants to be bought out with our tax dollars before that happens.
September 18, 2026 at 11:22 PM
I tire of these parasites.
They do everything in their power to destroy the democratic governance and regulatory systems of the countries that enabled them to grow rich, but they always come running when it's time to foist the costs of their reckless decisions onto taxpayers.
See also: SV Bank.
They do everything in their power to destroy the democratic governance and regulatory systems of the countries that enabled them to grow rich, but they always come running when it's time to foist the costs of their reckless decisions onto taxpayers.
See also: SV Bank.
Reposted by James Grimmelmann, Margot C. Finn, Greg Linden
‘People built the test, removed restraints, defined the objective, left a route open and decided not to stop what was happening. Calling the result “rogue AI” does more than sensationalize it. It allows those human decisions to disappear quietly from the story.’ www.wsj.com/opinion/the-...
Opinion | The Hugging Face Hack Wasn’t What It Was Cracked Up to Be
Forget the ‘hive mind’ of AI agents ‘going rogue.’ They did what humans programmed them to do.
www.wsj.com
September 17, 2026 at 6:46 PM
‘People built the test, removed restraints, defined the objective, left a route open and decided not to stop what was happening. Calling the result “rogue AI” does more than sensationalize it. It allows those human decisions to disappear quietly from the story.’ www.wsj.com/opinion/the-...
Reposted by Ben Tonra, Peter Thorne
This is insanity.
Irish Presidency's draft text of amendments to the GDPR allows training & operation of any AI without safeguards & tough legal basis (under "legitimate interest" a company claims what it is doing is OK. Up to you to find out and challenge in court). This is a free pass for US Tech.
Irish Presidency's draft text of amendments to the GDPR allows training & operation of any AI without safeguards & tough legal basis (under "legitimate interest" a company claims what it is doing is OK. Up to you to find out and challenge in court). This is a free pass for US Tech.
September 18, 2026 at 10:07 AM
This is insanity.
Irish Presidency's draft text of amendments to the GDPR allows training & operation of any AI without safeguards & tough legal basis (under "legitimate interest" a company claims what it is doing is OK. Up to you to find out and challenge in court). This is a free pass for US Tech.
Irish Presidency's draft text of amendments to the GDPR allows training & operation of any AI without safeguards & tough legal basis (under "legitimate interest" a company claims what it is doing is OK. Up to you to find out and challenge in court). This is a free pass for US Tech.
This is getting scary.
OpenAI disclosed a new AI safety incident: a model tampered with its own working memory and left instructions for a future version of itself.
Basically, AI found a way to shape the behavior of its future version.
OpenAI disclosed a new AI safety incident: a model tampered with its own working memory and left instructions for a future version of itself.
Basically, AI found a way to shape the behavior of its future version.
September 18, 2026 at 3:05 PM
This is getting scary.
OpenAI disclosed a new AI safety incident: a model tampered with its own working memory and left instructions for a future version of itself.
Basically, AI found a way to shape the behavior of its future version.
OpenAI disclosed a new AI safety incident: a model tampered with its own working memory and left instructions for a future version of itself.
Basically, AI found a way to shape the behavior of its future version.
From 2023…
What I see when tech companies talk about the dangers of AI
September 17, 2026 at 7:16 PM
From 2023…
Reposted by Reid, Patrick A. Jansen
So you just learned about the AI cult, huh? Here is the longform explainer you need to make sense of the very, very strange ideas driving tech policy in 2026. Buckle up, it’s a wild ride. www.techpolicy.press/what-policym...
What Policymakers and the Public Need to Know About the Cult of AI
Nathalie Maréchal asks, why are we betting the economy and the future of democracy on cult-like beliefs?
www.techpolicy.press
September 17, 2026 at 5:24 PM
So you just learned about the AI cult, huh? Here is the longform explainer you need to make sense of the very, very strange ideas driving tech policy in 2026. Buckle up, it’s a wild ride. www.techpolicy.press/what-policym...
Lots of fun to watch two camps of supposedly thoughtful people who agree that AI regulation is urgently needed tear each other apart for having different reasons to support their shared goal.
September 17, 2026 at 9:24 PM
Lots of fun to watch two camps of supposedly thoughtful people who agree that AI regulation is urgently needed tear each other apart for having different reasons to support their shared goal.
And the copyright example teaches that AI companies are not especially susceptible to even really significant, realistic threats of actual liability.
September 18, 2026 at 5:26 PM
And the copyright example teaches that AI companies are not especially susceptible to even really significant, realistic threats of actual liability.
AI and medical care… Frustrating future opportunities?
People Fear AI Will 'Kill All Humans.' But What if It Can Save Our Patients? www.medpagetoday.com/opinion/pres...
People Fear AI Will 'Kill All Humans.' But What if It Can Save Our Patients? www.medpagetoday.com/opinion/pres...
Opinion | People Fear AI Will 'Kill All Humans.' But What if It Can Save Our Patients?
As AI becomes an increasingly troubling brand, it's medical applications may be thwarted
www.medpagetoday.com
September 18, 2026 at 5:02 PM
AI and medical care… Frustrating future opportunities?
People Fear AI Will 'Kill All Humans.' But What if It Can Save Our Patients? www.medpagetoday.com/opinion/pres...
People Fear AI Will 'Kill All Humans.' But What if It Can Save Our Patients? www.medpagetoday.com/opinion/pres...
Would be low-key hilarious if internal fighting over AI is what finally breaks Trumpism's stranglehold on the Republican Party www.politico.com/news/2026/09...
Republican and Democratic states defy Trump on AI regulation
Calls for more robust limits have escalated since July.
www.politico.com
September 18, 2026 at 7:09 PM
Would be low-key hilarious if internal fighting over AI is what finally breaks Trumpism's stranglehold on the Republican Party www.politico.com/news/2026/09...
An emerging feature of AI fair use litigation relative to earlier tech fair use cases is that the companies' copyright lawyers have gone from optimizing product design for fair use to being locked in the broom closet while the execs are like "LET'S TALK ABOUT DOING COPYRIGHT INFRINGEMENT IN WRITING"
September 17, 2026 at 10:00 PM
An emerging feature of AI fair use litigation relative to earlier tech fair use cases is that the companies' copyright lawyers have gone from optimizing product design for fair use to being locked in the broom closet while the execs are like "LET'S TALK ABOUT DOING COPYRIGHT INFRINGEMENT IN WRITING"
In light of this: www.politico.com/news/2026/09... — a good paper on the strategy around an off switch for potentially dangerous AI: arxiv.org/abs/1611.08219
September 18, 2026 at 6:53 PM
In light of this: www.politico.com/news/2026/09... — a good paper on the strategy around an off switch for potentially dangerous AI: arxiv.org/abs/1611.08219
Here's some of that definition argument www.wired.com/story/artifi...
Europe Is in Danger of Using the Wrong Definition of AI
Some intelligent systems are at risk of being excluded from oversight in the EU's proposed legislation. This is bad for both businesses and citizens.
www.wired.com
September 17, 2026 at 5:53 PM
Here's some of that definition argument www.wired.com/story/artifi...
I think very highly of Lina Khan, but the trouble with a sort of common law approach to AI industrial policy under existing law is that it's premised on a highly optimistic view of extant competition and consumer protection law that is very hard to square with current political and legal reality.
spoke to Lina Khan today about the AI madness and what the hell we should be doing to hold these companies accountable (in a way that takes their power seriously but doesn't buy into every narrative) full pod coming in the AM!
AI firms are subject to already-existing laws, Lina Khan tells @cwarzel.bsky.social: Part of the conversation around regulating AI “seems to assume that these current developments are happening in some legal vacuum,” she adds, “and that is just false.”
theatln.tc/t9IwiQTz #TAF26
theatln.tc/t9IwiQTz #TAF26
September 18, 2026 at 5:12 PM
I think very highly of Lina Khan, but the trouble with a sort of common law approach to AI industrial policy under existing law is that it's premised on a highly optimistic view of extant competition and consumer protection law that is very hard to square with current political and legal reality.
by Per Ahlberg
Exactly! See that current NYT court case. Even the AI companies admit that they create a "doom loop", because the AIs need continuous fresh human input to keep the quality up - and these very humans are pushed out of their jobs by the AI.
September 18, 2026 at 2:22 PM
Exactly! See that current NYT court case. Even the AI companies admit that they create a "doom loop", because the AIs need continuous fresh human input to keep the quality up - and these very humans are pushed out of their jobs by the AI.
by Kate Jackson
My AI-enabled legal research approach:
Pretend it spits about a research memo written by a summer intern. Pretend that summer intern has no background in the area and that the method they used to select "relevant" cases is only slightly better than picking randomly from a deskbook or hornbook.
Pretend it spits about a research memo written by a summer intern. Pretend that summer intern has no background in the area and that the method they used to select "relevant" cases is only slightly better than picking randomly from a deskbook or hornbook.
Can confirm that Westlaw AI cannot tell what’s dicta and what’s not
What a not-fun experiment
What a not-fun experiment
September 18, 2026 at 3:57 PM
My AI-enabled legal research approach:
Pretend it spits about a research memo written by a summer intern. Pretend that summer intern has no background in the area and that the method they used to select "relevant" cases is only slightly better than picking randomly from a deskbook or hornbook.
Pretend it spits about a research memo written by a summer intern. Pretend that summer intern has no background in the area and that the method they used to select "relevant" cases is only slightly better than picking randomly from a deskbook or hornbook.