Anthropic says AI agents didn’t breach Australian government websites – video
TL;DR
At a joint parliamentary hearing on artificial intelligence in Australia, Anthropic's head of safeguards, Dave Orr, said a review of hundreds of millions of transcripts found no unauthorised interactions with Australian government systems. Orr acknowledged, however, that Anthropic has limited visibility into how customers use its models because of standard zero data retention policies. Anthropic also told the inquiry it never tried to dictate Australia's copyright rules.
Nauti's Take
Reviewing hundreds of millions of transcripts shows real potential: providers can audit misuse systematically and demonstrate it transparently to lawmakers. Anthropic named the limit in the same breath, since zero data retention leaves much of what customers do with the models out of view.
Governments and businesses should therefore keep their own logs instead of relying solely on provider oversight.
Sources
