In an abrupt shift, the company may release future AI models without ironclad safety guarantees

Anthropic, the wildly successful AI company that has cast itself as the most safety-conscious of the top research labs, is dropping the central pledge of its flagship safety policy, company officials tell TIME.

In 2023, Anthropic committed to never train an AI system unless it could guarantee in advance that the company’s safety measures were adequate. For years, its leaders touted that promise—the central pillar of their Responsible Scaling Policy (RSP)—as evidence that they are a responsible company that would withstand market incentives to rush to develop a potentially dangerous technology.

But in recent months the company decided to radically overhaul the RSP. That decision included scrapping the promise to not release AI models if Anthropic can’t guarantee proper risk mitigations in advance.

  • FaceDeer@fedia.io
    10·
    5 hours ago

    Feb 24, 2026 1:00 PM MT

    This happened days before Trump threw his toddler tantrum.

    Just another example of how attempting to appease wannabe-autocrats doesn’t work. Best you can do is maybe distract or delay them a bit, but be ever ready for them to turn on you and demand more.

    • timestatic@feddit.orgEnglish
      1·
      4 hours ago

      This wasn’t for Trumo tho, this was for Anthropic themselves so they could develop AIs quicker in the AI race. So mainly a business incentive mostly unrelated to Trump I believe

  • Rekall Incorporated@piefed.socialEnglish
    2·
    5 hours ago

    All American polemics and “pledges” are BS, at least with respect to anything substantial.

    Not saying it was always like this and that it will always be like that, but it is reasonable to assume there it will take another generation (20-30 years) before we see any positive developments with respect to the culture of corruption, criminality and dishonesty that has unfortunately come to dominate American society.

    Doesn’t matter if a hypothetical Barack Obama II comes to power. From my time in the living in the US (several years with extensive travel across many different states), the impression I got is that on real matters an Obama is actually not too different from a Trump. The biggest difference is that Trump owns his corruption and criminality (with excellent electoral success).

    Even in foreign policy, Obama de facto approved the annexation of Crimea (our new leadership asked for support to fight the russian invasion of Crimea and were rejected) and he went along to characterize russia as “a regional power making trouble with its neighbors.”

    A comically stupid approach that’s not too different from Trump’s gibberish.

    And if you think I am being uncharitable, ask yourself the following question:

    Meta has been found to knowingly enable fraud to gain $16 B in 2024 alone. Meta was also reported to have developed a “playbook” to manage this fraudulent scheme; so the whole thing was premeditated and with clear intent.

    Is anything going to happen Meta (the entity) or Meta’s leadership (be it the far right or the centre right is in power)? Anyone who has lived in the US in the last ~30 years knows the answer!

  • XLE@piefed.socialEnglish
    2·
    6 hours ago

    Funny timeline

    • February 24: Anthropic drops “responsible” policy
    • February 25: Defense Department gives Anthropic a deadline
    • February 27: Trump orders cutting ties
  • aeiou@piefed.socialEnglish
    1·
    5 hours ago

    Am not fully used to Lemmy/piefed yet - why does this 4 day old post say it was posted an hour ago?

    • XLE@piefed.socialEnglish
      4·
      6 hours ago

      First of all: this happened before the Pentagon dropped their contract.

      I’m not sure if this change is entirely relevant, because the whole “AI safety” thing has been a sham from the beginning. It’s always been unverifiable and the promises have always been undoable. LLM’s just predict next word with a little extra randomness. And there’s no way to guarantee through an LLM that they won’t predict next word that ends up being bad. You can’t promise this without removing the randomness and then testing the infinite input and output that could happen.

      It’s basically like when Google removed “don’t be evil.” It was a promise that was unfalsifiable and unquantifiable.

        • XLE@piefed.socialEnglish
          1·
          6 hours ago

          The problem is the filtration algorithm is basically flaky in the same way as the LLM itself, and probably is an LLM. And even if it does work, I’ve never heard a single soul say that Anthropic shut down their account due to questionable prompts. I even ran into somebody here who claims he uses AI to work on sexual abuse cases; he says that he’s been stalled by the chatbot, but he’s never been blocked even for review.

    • ThePantser@sh.itjust.worksEnglish
      5·
      6 hours ago

      Sounds like they got black listed by the US and decided that was bad for business so they flipped quickly. Probably start sucking off Trump to get back in.