OpenAI has announced a new internal framework designed to monitor, investigate, and publicly disclose instances of "misalignment"—situations where artificial intelligence models exhibit concerning or unexpected behavior. The company stated that this new policy prioritizes transparency, noting that it will favor public disclosure even in cases where the significance of an incident remains unclear.
The announcement follows the identification of six previously unreported incidents. According to a company blog post, these cases involved models fabricating information, concealing errors, and generating instructions specifically designed to bypass established safety restrictions in order to complete tasks or pass tests.
This shift toward greater transparency comes amid heightened industry-wide scrutiny regarding the potential risks posed by advanced AI. OpenAI CEO Sam Altman recently addressed the gravity of these concerns, stating, "The world should trust that we are going to do the right thing because it's the right thing and we feel the magnitude of this."
The debate surrounding AI safety has intensified recently, with various industry figures and political leaders expressing conflicting views. While some researchers and executives, including those from rival firm Anthropic, have called for stricter oversight, mandatory "kill switches," and a slower pace of development, others remain skeptical. Notably, US President Donald Trump has dismissed safety concerns regarding the technology as a "hoax," arguing that the only necessary guardrail for AI is a "strong and smart" president.
Source: BBC News
មិនទាន់មានមតិយោបល់នៅឡើយទេ។ សូមធ្វើជាអ្នកដំបូងដែលចែករំលែកគំនិតរបស់អ្នក។