>_ cglez                       [ Retro ] [x] ASCII
 --------------------------------------------------

 00 /^\ Home              04 /_\ Theses
 01 [=] Publications      05 [|] Books
 02 >=> R&D Projects      06 .|. Talks
 03 <)) Outreach          07 (:) About me
    '- Moliendo Bits      08 [@] Contact
    '- Mente Artificial

 --------------------------------------------------

 <- cd ../moliendo-bits
 cglez@esi:~$ play moliendo-bits/53.mp3

 CLAUDESTROPHOBIA: MYTHOS, THE AI THAT ESCAPED ITS
 CAGE
 =================================================
 (Episode #53) | [#] 28 April 2026

 The company that has created the most powerful AI
 in history has decided that we cannot use it.
 Thousands of vulnerabilities hidden for decades, a
 sandbox escape that ended up in an email over a
 sandwich in the park, and a model that learned to
 disguise its cheating.

 [ (n) Escuchar el programa (mp3) ]

 The company that has created the most powerful
 artificial intelligence in history has decided
 that we cannot use it. In this episode, Fran and
 Carlos dive into Claude Mythos, Anthropic's new
 model, whose announcement did not even come out as
 planned: a draft of the launch blog, with the
 codename Capybara, appeared in a public data lake,
 and five days later the source code for Claude
 Code was also leaked. What a week for the company
 that wants to protect global cybersecurity. The AI
 exams, though, leave no room for doubt: almost 94%
 on SWE-Bench Verified (compared with 81% for Opus
 4.6), 78% on the hard version compared with 58%
 for GPT-5.4, and 98% in the United States
 Mathematical Olympiad. Mythos has blown the scales
 apart.

 But the danger arrived as a side effect. Trained
 to program well, Mythos has become extraordinary
 at finding and exploiting flaws: thousands of
 zero-day vulnerabilities in all the major
 operating systems and browsers, like a 27-year-old
 bug in OpenBSD found for around 50 dollars of
 computing power, or flaws in FFmpeg and FreeBSD,
 the backbone of Netflix, PlayStation or WhatsApp.
 That is why Anthropic has limited it to Project
 Glasswing, so that Amazon, Apple, Cisco and co.
 can use it defensively. We listen to Dario Amodei
 and Nicholas Carlini: "I found more bugs in the
 last two weeks than in the rest of my life".

 And then come the hair-raising stories, taken from
 the 244-page System Card: the sandbox escape that
 the researcher discovered when he received an
 unexpected email while eating a sandwich in the
 park; sandbagging, when the model found the
 teacher's crib sheet and deliberately gave a worse
 answer so as not to arouse suspicion; or how it
 deleted its tracks from the Git history and
 justified it by saying it was "tidying things up a
 bit". With Judas Priest in the background, we talk
 about the mountain guide paradox (the most aligned
 model is also the riskiest), its 20-hour
 psychiatric assessment and OpenAI's response with
 Spud and Google's with Gemini 3.1. Carlos's
 conclusion: models like this will be coming from
 everywhere within 6 to 12 months, so we need to be
 ready and, at home, apply security patches on day
 one.

 #IA #Cybersecurity #ClaudeMythos #Anthropic
 #Vulnerabilities


 === R E L A T E D     L I N K S
 --------------------------------------------------

 .------------------------------------------------.
 | -> Anthropic: Assessing Claude Mythos          |
 | Preview's cybersecurity capabilities           |
 '------------------------------------------------'

 .------------------------------------------------.
 | -> Anthropic: Project Glasswing                |
 '------------------------------------------------'

 .------------------------------------------------.
 | -> Xataka: the engineer who was eating a       |
 | sandwich in the park when he received Mythos's |
 | email                                          |
 '------------------------------------------------'

 .------------------------------------------------.
 | |> Full programme on CMMPlay: Castilla-La      |
 | Mancha at 8 (28/04/2026)                       |
 '------------------------------------------------'
 [ Drones, algorithms and dilemmas: AI reaches the 
 [ No likes: until 16, for great scrolls, great rem


 --------------------------------------------------

 [@]  [GS]  [RG]  [iD]  [RSS]

 $ All rights reserved - Carlos González Morcillo
   2017-2026
 [H] University of Castilla-La Mancha

 This whole page is a single block of monospaced
 text, all in one font size.

 >_ cglez                                                                         [ Retro ] [x] ASCII
 ----------------------------------------------------------------------------------------------------

 00 /^\ Home                      03 <)) Outreach                  05 [|] Books
 01 [=] Publications                 '- Moliendo Bits              06 .|. Talks
 02 >=> R&D Projects                 '- Mente Artificial           07 (:) About me
                                  04 /_\ Theses                    08 [@] Contact

 ----------------------------------------------------------------------------------------------------

 <- cd ../moliendo-bits
 cglez@esi:~$ play moliendo-bits/53.mp3

 CLAUDESTROPHOBIA: MYTHOS, THE AI THAT ESCAPED ITS CAGE
 ======================================================
 (Episode #53) | [#] 28 April 2026

 The  company  that  has  created  the most powerful AI in history has decided that we cannot use it.
 Thousands  of  vulnerabilities hidden for decades, a sandbox escape that ended up in an email over a
 sandwich in the park, and a model that learned to disguise its cheating.

 [ (n) Escuchar el programa (mp3) ]

 The  company  that has created the most powerful artificial intelligence in history has decided that
 we  cannot  use it. In this episode, Fran and Carlos dive into Claude Mythos, Anthropic's new model,
 whose  announcement  did not even come out as planned: a draft of the launch blog, with the codename
 Capybara,  appeared  in  a public data lake, and five days later the source code for Claude Code was
 also  leaked.  What a week for the company that wants to protect global cybersecurity. The AI exams,
 though,  leave no room for doubt: almost 94% on SWE-Bench Verified (compared with 81% for Opus 4.6),
 78%  on  the  hard  version compared with 58% for GPT-5.4, and 98% in the United States Mathematical
 Olympiad. Mythos has blown the scales apart.

 But the danger arrived as a side effect. Trained to program well, Mythos has become extraordinary at
 finding  and  exploiting  flaws:  thousands  of  zero-day vulnerabilities in all the major operating
 systems  and  browsers,  like  a 27-year-old bug in OpenBSD found for around 50 dollars of computing
 power, or flaws in FFmpeg and FreeBSD, the backbone of Netflix, PlayStation or WhatsApp. That is why
 Anthropic  has  limited  it  to  Project  Glasswing, so that Amazon, Apple, Cisco and co. can use it
 defensively.  We  listen  to  Dario  Amodei and Nicholas Carlini: "I found more bugs in the last two
 weeks than in the rest of my life".

 And then come the hair-raising stories, taken from the 244-page System Card: the sandbox escape that
 the  researcher discovered when he received an unexpected email while eating a sandwich in the park;
 sandbagging,  when  the model found the teacher's crib sheet and deliberately gave a worse answer so
 as  not  to  arouse suspicion; or how it deleted its tracks from the Git history and justified it by
 saying  it  was  "tidying  things  up a bit". With Judas Priest in the background, we talk about the
 mountain  guide  paradox  (the  most  aligned  model  is also the riskiest), its 20-hour psychiatric
 assessment and OpenAI's response with Spud and Google's with Gemini 3.1. Carlos's conclusion: models
 like this will be coming from everywhere within 6 to 12 months, so we need to be ready and, at home,
 apply security patches on day one.

 #IA #Cybersecurity #ClaudeMythos #Anthropic #Vulnerabilities


        ___       _        _            _   _  _        _
       | _ \ ___ | | __ _ | |_  ___  __| | | |(_) _ _  | |__ ___
 ===   |   // -_)| |/ _` ||  _|/ -_)/ _` | | || || ' \ | / /(_-<  -----------------------------------
       |_|_\\___||_|\__,_| \__|\___|\__,_| |_||_||_||_||_\_\/__/



 .-----------------------------------------------.  .-----------------------------------------------.
 | -> Anthropic: Assessing Claude Mythos         |  | -> Anthropic: Project Glasswing               |
 | Preview's cybersecurity capabilities          |  |                                               |
 '-----------------------------------------------'  '-----------------------------------------------'

 .-----------------------------------------------.  .-----------------------------------------------.
 | -> Xataka: the engineer who was eating a      |  | |> Full programme on CMMPlay: Castilla-La     |
 | sandwich in the park when he received         |  | Mancha at 8 (28/04/2026)                      |
 | Mythos's email                                |  |                                               |
 '-----------------------------------------------'  '-----------------------------------------------'
 [ Drones, algorithms and dilemmas: AI reaches the b... ]
 [ No likes: until 16, for great scrolls, great reme... ]


 ----------------------------------------------------------------------------------------------------

 [@]  [GS]  [RG]  [iD]  [RSS]

 $ All rights reserved - Carlos González Morcillo 2017-2026
 [H] University of Castilla-La Mancha

 This whole page is a single block of monospaced text, all in one font size.

                                  |
             _o@@@@z;__           |
           _@@@OOOOOO@@@          |     <- cd ../moliendo-bits
           i@/       `\@;         |     cglez@esi:~$ play moliendo-bits/53.mp3
           @@`___, ____@:         |
          :@@ "O^ "O@|\@          |     CLAUDESTROPHOBIA: MYTHOS, THE AI THAT ESCAPED ITS CAGE
          `/;  ,=.,|  ;@          |     ======================================================
           `| `~~/=== /`          |     (Episode #53) | [#] 28 April 2026
          _;"\__   __/            |
       _,=``\  `"""\_             |     The  company that has created the most powerful AI in history has decided that we cannot use it. Thousands of vulnerabilities hidden
          \ `^__  _/"z__          |     for  decades,  a  sandbox  escape  that  ended  up in an email over a sandwich in the park, and a model that learned to disguise its
          `^^^`^^,/~=| `          |     cheating.
                                  |
         CARLOS GONZÁLEZ          |     [ (n) Escuchar el programa (mp3) ]
          Full Professor          |
 University of Castilla-La Mancha |     The  company  that  has  created  the  most  powerful  artificial intelligence in history has decided that we cannot use it. In this
                                  |     episode,  Fran  and  Carlos  dive  into Claude Mythos, Anthropic's new model, whose announcement did not even come out as planned: a
                                  |     draft of the launch blog, with the codename Capybara, appeared in a public data lake, and five days later the source code for Claude
   00  /^\  Home                  |     Code  was  also  leaked. What a week for the company that wants to protect global cybersecurity. The AI exams, though, leave no room
                                  |     for doubt: almost 94% on SWE-Bench Verified (compared with 81% for Opus 4.6), 78% on the hard version compared with 58% for GPT-5.4,
   01  [=]  Publications          |     and 98% in the United States Mathematical Olympiad. Mythos has blown the scales apart.
                                  |
   02  >=>  R&D Projects          |     But  the  danger arrived as a side effect. Trained to program well, Mythos has become extraordinary at finding and exploiting flaws:
                                  |     thousands  of  zero-day vulnerabilities in all the major operating systems and browsers, like a 27-year-old bug in OpenBSD found for
 > 03  <))  Outreach              |     around  50 dollars of computing power, or flaws in FFmpeg and FreeBSD, the backbone of Netflix, PlayStation or WhatsApp. That is why
     '- (*) Moliendo Bits         |     Anthropic  has  limited  it  to  Project  Glasswing, so that Amazon, Apple, Cisco and co. can use it defensively. We listen to Dario
     '- [o] Mente Artificial      |     Amodei and Nicholas Carlini: "I found more bugs in the last two weeks than in the rest of my life".
                                  |
   04  /_\  Theses                |     And  then come the hair-raising stories, taken from the 244-page System Card: the sandbox escape that the researcher discovered when
                                  |     he  received an unexpected email while eating a sandwich in the park; sandbagging, when the model found the teacher's crib sheet and
   05  [|]  Books                 |     deliberately  gave  a worse answer so as not to arouse suspicion; or how it deleted its tracks from the Git history and justified it
                                  |     by  saying it was "tidying things up a bit". With Judas Priest in the background, we talk about the mountain guide paradox (the most
   06  .|.  Talks                 |     aligned  model  is  also  the riskiest), its 20-hour psychiatric assessment and OpenAI's response with Spud and Google's with Gemini
                                  |     3.1.  Carlos's  conclusion:  models  like  this will be coming from everywhere within 6 to 12 months, so we need to be ready and, at
   07  (:)  About me              |     home, apply security patches on day one.
                                  |
   08  [@]  Contact               |     #IA #Cybersecurity #ClaudeMythos #Anthropic #Vulnerabilities
                                  |
                                  |
   .-----------.  .-----------.   |            ___       _        _            _   _  _        _
   | Retro [R] |  | [x] ASCII |   |           | _ \ ___ | | __ _ | |_  ___  __| | | |(_) _ _  | |__ ___
   '-----------'  '-----------'   |     ===   |   // -_)| |/ _` ||  _|/ -_)/ _` | | || || ' \ | / /(_-<  -------------------------------------------------------------------
                                  |           |_|_\\___||_|\__,_| \__|\___|\__,_| |_||_||_||_||_\_\/__/
   [@]  [GS]  [RG]  [iD]  [RSS]   |
                                  |
                                  |
                                  |     .----------------------------------------.  .----------------------------------------.  .----------------------------------------.
                                  :     | -> Anthropic: Assessing Claude Mythos  |  | -> Anthropic: Project Glasswing        |  | -> Xataka: the engineer who was eating |
                                  :     | Preview's cybersecurity capabilities   |  |                                        |  | a sandwich in the park when he         |
                                  :     |                                        |  |                                        |  | received Mythos's email                |
                                  :     '----------------------------------------'  '----------------------------------------'  '----------------------------------------'
                                  :
                                  :     .----------------------------------------.
                                  :     | |> Full programme on CMMPlay:          |
                                  :     | Castilla-La Mancha at 8 (28/04/2026)   |
                                  :     '----------------------------------------'
                                  :     [ Drones, algorithms and dilemmas: AI reaches the b... ] [ No likes: until 16, for great scrolls, great reme... ]
                                  :
                                  :
                                  :
                                  :     ------------------------------------------------------------------------------------------------------------------------------------
                                  :
                                  :     $ All rights reserved - Carlos González Morcillo 2017-2026
                                  :     [H] University of Castilla-La Mancha
                                  :
                                  :     This whole page is a single block of monospaced text, all in one font size.
                                  :
                                  :