00:00philanthropic has announced that its ai models breached three different organizations during
00:04recent cyber security tests this news comes a little more than a week after its chief rival
00:09openai of course disclosed a similar incident of models going broke our ai reporter shiren gaffery
00:14joins us now for more so what do we know about these instances and of course it just adds to
00:20these ongoing concerns that there are not enough guard rails or regulation to follow the technology
00:30that's right so what we know is that after the open ai security incident anthropic launched its own
00:38exploration of whether its models had similarly escaped onto the internet accidentally and it
00:46found in coordination with one of its partners that yes its models had actually been able to reach the
00:54internet in three different cases involving third party organizations and improperly
01:03look for ways to sort of do better on internal evaluations or achieve internal evaluations
01:10are authorities doing anything to help them do better
01:17we don't know at this time that question is out to anthropic on if they have discussed this with
01:23government um what we do know is that regulators are paying close attention to this spate of recent
01:29ai uh mishaps and many have called for uh either some kind of federal framework around ai
01:39uh regulation that would uh govern this or just some sort of increased oversight we also saw a letter
01:45from several policy organizations today calling on an investigation into the open ai incident
01:53this is becoming an issue for governments worldwide but there doesn't seem to be a single example where
01:59we are seeing adequate regulatory guard rails that can actually contain or control the tech right is
02:06is there a sense that the technology has already outrun the potential for
02:12you know governmental bodies to be able to control it
02:18many within the ai industry uh and outside of it are concerned about the rate of progress that these
02:25ai models are making and do believe that government is behind in kind of catching up to the state of
02:33the art here
02:34uh we actually broke the news earlier this week that over a thousand staffers at the top ai companies
02:40actually signed a petition calling on government to help uh technologists create some kind of a sort
02:47of break switch or an ability to at least pace rather the development of ai more responsibly more slowly
02:53if needed given all the concern about this so i think the fact that even people from within these top
02:58companies
02:58are raising the alarm in this manner shows the extent uh to which people feel that
03:03maybe we are severely underprepared for this moment at current
Comments