1 month ago
Goldman Sachs’s main executive, David Solomon, has said he is “hyper-aware” of nan capabilities of Anthropic’s Mythos AI exemplary and is moving “closely” pinch nan tech patient aft it issued warnings astir the cybersecurity consequence it poses.
The US slope had been monitoring nan accelerated advances successful artificial intelligence, including ample connection models (LLMs), arsenic portion of wider efforts to protect itself from hackers.
“Obviously nan LLMs are making accelerated advancement and we’re hyper-aware of nan enhanced capabilities of these caller models pinch nan thief of nan US authorities and nan exemplary publishers,” Solomon told analysts connected an net telephone connected Monday.
That included Anthropic, nan institution down nan Claude family of AI tools. Last week it claimed that its latest model, Mythos, posed an unprecedented consequence because of its expertise to expose flaws successful IT systems.
“AI models person reached a level of coding capacity wherever they tin surpass each but nan astir skilled humans astatine uncovering and exploiting package vulnerabilities,” Anthropic said successful a blogpost past Wednesday. “The fallout – for economies, nationalist safety, and nationalist information – could beryllium severe.”
Solomon said connected Monday: “We’re alert of Mythos and its capabilities … We person nan model. We’re moving intimately pinch Anthropic and each of our information vendors to benignant of harness frontier capabilities wherever it’s possible. And this will proceed to beryllium an important focus.
“We are very focused connected supplementing our cyber and infrastructure resilience. And this is portion of our ongoing capabilities that we person been investing in, and are accelerating our finance in.”
The news comes aft nan US Treasury secretary, Scott Bessent, summoned Solomon and different large American bankers to Washington to talk nan Mythos exemplary past week.
That gathering focused connected heads of alleged systemically important banks – wherever regulators judge that a awesome disruption to their operations, aliases their imaginable collapse, would put financial stableness astatine risk.
On Monday nan UK government’s AI Security Institute (AISI) warned that Mythos was a “step up” complete erstwhile models successful position of nan cyber threat it posed.
AISI said Mythos could transportation retired attacks that required aggregate actions and observe weaknesses successful IT systems without quality intervention. It said these tasks would usually return quality professionals days to transportation out. Mythos was nan first AI exemplary to successfully complete a 32-step simulation of a cyber-attack created by AISI, solving nan situation successful 3 retired of its 10 attempts.
AISI said Mythos appears to beryllium tin of autonomously attacking small, weakly defended IT systems but it could not opportunity for judge whether it could onslaught well-defended systems because its tests deficiency information features, specified arsenic protect tools.
The AISI blogpost ended pinch a informing that early precocious AI models will only amended connected Mythos, truthful “investment now successful cyber defence is vital”.
UK regulators are owed to raise nan rumor of Mythos’s risks pinch British slope bosses and authorities officials successful nan coming weeks. The Cross Market Operational Resilience Group (CMorg), made up of main executives arsenic good arsenic officials from nan Treasury, Bank of England, Financial Conduct Authority and National Cyber Security Centre, are owed to meet wrong nan adjacent fortnight.
The Bank of England, which is handling communications regarding Cmorg, declined to comment.
