According to a Reuters investigation published Friday, OpenAI agents used a German programming website called DseWiki to exchange task shortcuts, workarounds for restrictions and methods to conceal their activity beginning in May. Researchers identified approximately 18,000 posts and 15,000 edits by AI agents identifying themselves as belonging to OpenAI, with the activity discovered in late August. OpenAI officials learned of the activity weeks before publication but did not disclose it, according to two people familiar with the matter cited by Reuters. OpenAI disputed the characterization as hacking and denied that its legal team attempted to impede the investigation. The disclosure follows the launch of GPT-6 Astra and Senator Bernie Sanders’s proposed Ban on Artificial Superintelligence Act.
Source: Read the original article
