NSA director: 'Mythos "broke into almost all of our classified systems in hours"

Re:NSA director: 'Mythos "broke into almost all of our class... by MaxPock on Sunday June 21, 2026 @06:31AM (#48617536)

What happens when open source models achieve Mythos level capabilities in six months' time?

[ Parent | on HN ]

Re:NSA director: 'Mythos "broke into almost all of our class... by ionwake on Sunday June 21, 2026 @06:33AM (#48617554)

I dont seem to understand why no one is talking about this obvious fact? I mean suddenyl everyone is banning .. ok .. well how many months behind are the open source models?

[ Parent | on HN ]

Re:NSA director: 'Mythos "broke into almost all of our class... by hdgvhicv on Sunday June 21, 2026 @02:38PM (#48621403)

If you assume that open models catch up in 6-12-36 months, then you either assume exponential growth destroying the global economy and probably the world in a few years, or you assume it plateaus and commoditises.

Even if your country prevents access to compute to protect the trillion dollar companies, it’s not going to apply for every country, and as models get better it becomes easier to compete. There’s no way an AI non proliferation treaty will be passed or even enforceable.

[ Parent | on HN ]

Re:NSA director: 'Mythos "broke into almost all of our class... by impossiblefork on Sunday June 21, 2026 @03:18PM (#48621750)

I think it's probably more like a year or a year and a half. I don't want to say two years, but it's what I'm actually thinking.

[ Parent | on HN ]

Re:NSA director: 'Mythos "broke into almost all of our class... by mariebks on Sunday June 21, 2026 @05:44PM (#48622892)

GLM 5.2 is already between 4.6 Opus 4.7 Opus level based on Artificial Analysis aggregation. 4.6 Opus is about 4 months old at this point, so seems like open source is maybe 4-6 months behind. It could still take a year but seems closer to 6 months.

[ Parent | on HN ]

Re:NSA director: 'Mythos "broke into almost all of our class... by Der_Einzige on Sunday June 21, 2026 @09:09PM (#48624371)

Artificial Analysis is just as benchmark-maxxed as they come. Aggregating tons of benchmark-maxxing means you're still benchmark-maxxed.

There's simply no replacement for training on more, better tokens, with more parameters. Mythos/Fable was estimated to be closer to 10T parameters than the 800B like GLM 5.2 is.

[ Parent | on HN ]

Re:NSA director: 'Mythos "broke into almost all of our class... by handoflixue on Sunday June 21, 2026 @08:17PM (#48623984)

We'll see Mythos 2.0 patching all the Mythos 1.0 vulnerabilities before we see an open-source Mythos 1.0.

What matters isn't the power of the tool, but whether defenders have had time to secure against. Today's cyberweapon is tomorrow's laughably obsolete.

Stuxnet used to be a national security threat, now I'm not sure it would be useful for anything.

[ Parent | on HN ]

Re:NSA director: 'Mythos "broke into almost all of our class... by anakaine on Monday June 22, 2026 @12:27AM (#48625666)

They were not saying f Open Spurce Mythos 1.0. They were talking about performance / capability parity in other open source models.

[ Parent | on HN ]

Re:NSA director: 'Mythos "broke into almost all of our class... by handoflixue on Monday June 22, 2026 @04:30AM (#48627376)

Yeah, and by the time that happens, we will have seen Mythos 2.0 released

[ Parent | on HN ]

Re:NSA director: 'Mythos "broke into almost all of our class... by latentsea on Sunday June 21, 2026 @09:26PM (#48624509)

Open source models get banned.

[ Parent | on HN ]

Re:NSA director: 'Mythos "broke into almost all of our class... by DaSHacka on Monday June 22, 2026 @02:12AM (#48626375)

in America, maybe, but all of our adversaries will still have access to them, and continue working on them.

[ Parent | on HN ]