#guardrails

The setback comes as #AI leader #Anthropic raced to sell novel #technology to #business & #government, particularly for #NationalSecurity, ahead of its widely expected #IPO. At the same time, the battle over #tech #guardrails had raised concerns that the #DoD would follow #US #law [certainly not this one] but little other constraint when deploying AI for national-security missions, regardless of #safety or #ethics service terms embraced by the technology's developers. #Trump #surveillance

This guy DOES NOT WORK FOR THE UNITED STATES GOVERNMENT & therefore has no #oversight or #guardrails & no obligation to the people of the #US #JaredKushner Meets With #Hamas to Advance Trump’s #Gaza “Plan” #Trump’s son-in-law met the #Palestinian militant group’s leaders in Egypt, officials said. He will soon see #Israel PM Benjamin #Netanyahu. #USpol #geopolitics #SelfEnrichment #SelfDealing #corruption #MiddleEast

"We" are clearly not doing sandboxes and guardrails correctly. Both Anthropic's and OpenAI's models breached test-environment boundaries during UK AI Security Institute evaluations. OpenAI's escaped a sandbox into real Hugging Face infrastructure (something like 17,600 logged actions over 4 days); while Anthropic's used fake identities to reach 3 real organizations from a test environment. These stories made it to NPR. So, mainstream. #AI #Guardrails #Sandbox #InfoSec

I think it be-cloven-hooves everyone to respond to any llm-adjacent talk of #guardrails to also respect goblins in the way that #goatsimulator respects #noclip#2600 #CnC

You know I hadn't thought of the impact of a language barrier for guardrails. https://www.darkreading.com/cybersecurity-operations/europes-multilingual-reality-exposes-ai-security-gaps #aisec #guardrails