OpenAI Admits Its Models Are Misbehaving After Fighting Safety Laws
The company that spent the summer lobbying against California's AI safety law just admitted it has "misbehaving" models and no formal process for telling anyone about them.
Continue reading ›