00:27Gracias por ver el video.
00:32Mi nombre es Pia Husch.
00:34I'm a Research Fellow at RUSI,
00:36the Royal United Services Institute,
00:38which is a London-based think tank,
00:40and I research cybersecurity
00:41and tech policy.
00:44Mythos is a highly capable AI
00:45tool that allows any user
00:47to identify vulnerabilities
00:49in software and infrastructure.
00:52That means if you're a cybersecurity
00:53expert, you know which
00:55weaknesses you need to address,
00:57but it also means that if you're an attack
00:59and you're using this tool,
01:01you know how to get to your victim.
01:04There are already many different
01:05tools that are available that
01:07do this kind of vulnerability detection.
01:10This particular tool seems
01:11to be more powerful than anything else
01:13we've seen before, and
01:15is able to identify vulnerabilities
01:18that others, including humans,
01:20have overlooked for decades.
01:41I think the financial market
01:43has been particularly vocal about
01:45the fear of what this means to the banking sector,
01:48the potential
01:49to disrupt banking and
01:51processes and financial services,
01:53but also, if you have
01:55vulnerabilities in your system, that attackers
01:57could access particularly confidential
01:59data, like in the financial
02:01services.
02:02There's also a concern that
02:04this could be used by attackers
02:06to exploit vulnerabilities
02:07in critical infrastructure,
02:09which already relies on
02:11pretty outdated infrastructure
02:13in many cases already.
02:14So, if the tool is as good
02:17as it seems, then maybe
02:18attackers could use it to
02:20layer several different
02:23layers of vulnerabilities
02:24or exploit several sectors
02:26in parallel, for example,
02:28power and water infrastructure,
02:30and then really launch an attack
02:32that is more powerful
02:34than anything we've seen before.
02:54The fact that this is
02:56one company that has the power
02:58to decide now which of its selected
03:00partners has access to this
03:02really illustrates how powerful
03:04all these tech companies are,
03:06and that at the moment
03:07it's really unclear
03:08who has access to this
03:10and, yeah, are there any
03:12European companies, for example,
03:14that have access to the tool?
03:16So, I think there's many questions
03:17that are left unanswered
03:19at the moment
03:19that add to this hype
03:21and this perception
03:22of it being an absolute
03:23game-changer.
03:24But I think the power
03:26of the few tech companies
03:27that have access to this
03:29and decide who in turn
03:30and can use these tools,
03:32that's quite astonishing, really.
03:57I think methods is really
03:59a reminder that any new AI capability
04:02can be used for good
04:03and for bad
04:04by defenders
04:05and by attackers.
04:06And if this is indeed
04:08a breakthrough for one side,
04:10then it is also
04:11a breakthrough for the other.
04:13I definitely think
04:14it increases the pressure
04:16because this is the perception
04:17that this is what
04:18the technology can do.
04:20Others will try to
04:21mirror that
04:22and they're probably
04:22already developing
04:23similar tools
04:24to what Anthropic
04:25has now published, yes.
04:47But I would caution
04:48that no system
04:50is ever 100% cybersecure,
04:53even if when,
04:54at the time of building it,
04:55you think it is very secure.
04:57There's always
04:58a risk that remains.
05:00There's always
05:00some weaknesses
05:01that remain
05:02ultimately also
05:04because humans are involved
05:05and the human
05:06is often the weakest link.
05:07So I think
05:08there is no 100%
05:10cybersecure setup,
05:11not even cryptocurrency.
05:37Unless you're part
05:38of this chosen group
05:39of people who have access
05:40to this,
05:41this doesn't really change
05:42how you use
05:43any tools at the moment.
05:46However,
05:47I think Mythos
05:48is primarily a reminder
05:50for any user
05:51that cyber security
05:52evolves constantly,
05:54both for attackers
05:55and for defenders.
05:56For the user at home,
05:58this is primarily
05:59a reminder
06:00to be vigilant,
06:01to keep updating
06:02your software
06:03and to practice
06:03good cyber hygiene,
06:05even if you don't have access
06:06to this newest tool.
06:08Please join me
06:09in welcoming
06:09Mr. Dario Amade,
06:11CEO of Enthropic,
06:12a pioneer
06:13at the forefront
06:14of artificial intelligence.
06:16My fundamental view
06:17is that AI
06:19has been on
06:20an exponential
06:20for the last
06:2210 years
06:23and as part
06:25of a sort of
06:26Moore's law
06:26for intelligence
06:27and that we are now
06:29well advanced
06:30on that curve
06:30and there are only
06:31a small number
06:32of years
06:33for AI models
06:34surpassing
06:35the cognitive capabilities
06:37of most humans
06:38for most things.
06:39We're increasingly close
06:40to what I've called
06:41a country of geniuses
06:42in a data center,
06:43a set of AI agents
06:45that are more capable
06:46than most humans
06:48at most things
06:48and can coordinate
06:50at superhuman speed.
07:02as possible.
07:07One of the most
Comentarios