Ethan Mollick (@emollick)

So the concern over Mythos and cybersecurity seems warranted.

2026-04-14 View on X

Politico

Sources: Anthropic largely left European regulators out of the loop as it limited Mythos' release to select companies and organizations; the UK's AISI tested it

BRUSSELS — Regulators in Europe have been left out of the loop as U.S. firm Anthropic restricts the release of a new, powerful artificial intelligence model.

View original

So the concern over Mythos and cybersecurity seems warranted.

2026-04-14 View on X

AI Security Institute

Cybersecurity analysis: Claude Mythos Preview had a 73% success rate on expert-level capture-the-flag challenges, which no model could finish before April 2025

one from Anthropic and another from an independent gov research group. …

View original

At this point, I assume that every internal memo at an AI lab is just written for public release. The labs are certainly capable of keeping secrets that don't leak, so they must realize that all-hands memos or announcements just go right to The Information or Wired or whatever.

2026-04-14 View on X

The Verge

Internal memo: OpenAI's Chief Revenue Officer says Anthropic is “grossing up rev share with Amazon and Google”, which overstates its “run rate by roughly $8B”

OpenAI's chief revenue officer, Denise Dresser, sent a four-page memo to employees on Sunday …

View original

So the concern over Claude Mythos and cybersecurity seems warranted based on this independent assessment from the UK government. It was capable of the equivalent of 20 hours of expert human work autonomously. — It is not an unexpected jump in capability, but it is big. www.aisi.gov.uk/blog/our-eva... …

2026-04-14 View on X

Politico

Sources: Anthropic largely left European regulators out of the loop as it limited Mythos' release to select companies and organizations; the UK's AISI tested it

BRUSSELS — Regulators in Europe have been left out of the loop as U.S. firm Anthropic restricts the release of a new, powerful artificial intelligence model.

View original

So the concern over Claude Mythos and cybersecurity seems warranted based on this independent assessment from the UK government. It was capable of the equivalent of 20 hours of expert human work autonomously. — It is not an unexpected jump in capability, but it is big. www.aisi.gov.uk/blog/our-eva... …

2026-04-14 View on X

AI Security Institute

Cybersecurity analysis: Claude Mythos Preview had a 73% success rate on expert-level capture-the-flag challenges, which no model could finish before April 2025

one from Anthropic and another from an independent gov research group. …

View original

At this point, I assume that every internal memo at an AI lab is just written for public release. The labs are certainly capable of keeping secrets that don't leak, so they must realize that all-hands memos or announcements just go right to The Information or Wired or whatever.

2026-04-13 View on X

The Verge

Internal memo: OpenAI Chief Revenue Officer says Anthropic is “grossing up rev share with Amazon and Google” and overstating its “run rate by roughly $8B”

OpenAI's chief revenue officer, Denise Dresser, sent a four-page memo to employees on Sunday …

View original

At this point, I assume that every internal memo at an AI lab is just written for public release. The labs are certainly capable of keeping secrets that don't leak, so they must realize that all-hands memos or announcements just go right to The Information or Wired or whatever.

2026-04-13 View on X

CNBC

Memo: OpenAI Chief Revenue Officer Denise Dresser says OpenAI's Microsoft deal “limited our ability” to reach clients using Bedrock and touts its Amazon deal

OpenAI's newly appointed revenue chief, Denise Dresser, sent a memo to staffers on Sunday, touting the company's alliance …

View original

I am catching glimpses in my feed that there is a backlash against Mythos as “marketing hype,” and it is a little confusing. I don't think anyone who has used the latest agentic coding tools, would think that expecting large-scale cybersecurity implications of increasingly good

2026-04-13 View on X

AI Security Institute

Cybersecurity analysis: Claude Mythos Preview had a 73% success rate on expert-level capture-the-flag challenges, which no model could finish before April 2025

The AI Security Institute (AISI) conducted evaluations of Anthropic's Claude Mythos Preview (announced on 7th April) to assess its cybersecurity capabilities.

View original

I am catching glimpses in my feed that there is a backlash against Mythos as “marketing hype,” and it is a little confusing. I don't think anyone who has used the latest agentic coding tools, would think that expecting large-scale cybersecurity implications of increasingly good

2026-04-13 View on X

Don't Worry About the Vase

A deep dive into the debate about Claude Mythos Preview, the model's capabilities, attempts to refute Anthropic's claims, and what it means for the future of AI

Anthropic is not going to release its new most capable model, Claude Mythos, to the public any time soon.

View original

So the concern over Claude Mythos and cybersecurity seems warranted based on this independent assessment from the UK government. It was capable of the equivalent of 20 hours of expert human work autonomously. — It is not an unexpected jump in capability, but it is big. www.aisi.gov.uk/blog/our-eva... …

2026-04-13 View on X

AI Security Institute

Cybersecurity analysis: Claude Mythos Preview had a 73% success rate on expert-level capture-the-flag challenges, which no model could finish before April 2025

The AI Security Institute (AISI) conducted evaluations of Anthropic's Claude Mythos Preview (announced on 7th April) to assess its cybersecurity capabilities.

View original

AI is jagged, but I think sometimes it is easy to overly focus on that. The generalness is a surprise too! LLMs may be optimized for verifiable fields like coding, but AI is also not bad at corporate strategy & medical advice & writing a sestina & expressing empathy & ideation.

2026-04-10 View on X

@karpathy

A gap in understanding AI is growing, as casual users cite flaws in old free models while power users point to new models' staggering gains in technical domains

Judging by my tl there is a growing gap in understanding of AI capability. The first issue I think is around recency and tier of use. I think a lot of people tried the free tier of...

View original

Seems like a good model from Meta that is still trailing the current series of releases. The most important thing to note is that it is not open weights. That was the main reason that Meta's models were so important. Without that, it is a lot harder to predict the value of Spark

2026-04-09 View on X

CNBC

Meta is opening a private API preview for Muse Spark to select partners, and plans to offer paid API access to a wider audience later; META closed up 6.5%

Jonathan Vanian /CNBC:

View original

Seems like a good model from Meta that is still trailing the current series of releases. The most important thing to note is that it is not open weights. That was the main reason that Meta's models were so important. Without that, it is a lot harder to predict the value of Spark

2026-04-09 View on X

Financial Times

Meta releases Muse Spark, the first model from Meta Superintelligence Labs under Alexandr Wang, to “power a smarter and faster” Meta AI across Meta's products

Muse Spark ‘purpose-built’ for social media apps as investors question huge AI investment

View original

I think the most obvious is that Meta has its own frontier model and can use that to extract additional value out of its customer base/explore new markets for its products. Very few companies can say that, and it has value on its own.

2026-04-09 View on X

Axios

Meta says Muse Spark powers queries in Meta AI and its “shopping mode” feature, and it plans to release a version of Muse Spark under an open-source license

Meta on Wednesday debuted Muse Spark, a homegrown AI model it says significantly narrows the performance gap with models from OpenAI, Anthropic and others.

View original

I think the most obvious is that Meta has its own frontier model and can use that to extract additional value out of its customer base/explore new markets for its products. Very few companies can say that, and it has value on its own.

2026-04-09 View on X

CNBC

Meta is opening a private API preview for Muse Spark to select partners, and plans to offer paid API access to a wider audience later; META closed up 6.5%

Jonathan Vanian /CNBC:

View original

I think the most obvious is that Meta has its own frontier model and can use that to extract additional value out of its customer base/explore new markets for its products. Very few companies can say that, and it has value on its own.

2026-04-09 View on X

Financial Times

Meta releases Muse Spark, the first model from Meta Superintelligence Labs under Alexandr Wang, to “power a smarter and faster” Meta AI across Meta's products

Muse Spark ‘purpose-built’ for social media apps as investors question huge AI investment

View original

Seems like a good model from Meta that is still trailing the current series of releases. The most important thing to note is that it is not open weights. That was the main reason that Meta's models were so important. Without that, it is a lot harder to predict the value of Spark

2026-04-09 View on X

Axios

Meta says Muse Spark powers queries in Meta AI and its “shopping mode” feature, and it plans to release a version of Muse Spark under an open-source license

Meta on Wednesday debuted Muse Spark, a homegrown AI model it says significantly narrows the performance gap with models from OpenAI, Anthropic and others.

View original

Seems like a good model from Meta that is still trailing the current series of releases. The most important thing to note is that it is not open weights. That was the main reason that Meta's models were so important. Without that, it is a lot harder to predict the value of Spark

2026-04-08 View on X

Financial Times

Meta releases Muse Spark, the first model from Meta Superintelligence Labs under Alexandr Wang, to “power a smarter and faster” Meta AI across Meta's products

Muse Spark ‘purpose-built’ for social media apps as investors question huge AI investment

View original

I was told about the Mythos release, but didn't have access, so have no personal experience to add. Two points from brief: 1) It is not built for IT security, it is just a good enough model that it is good at that too 2) This is the first, not last, model to raise security risks

2026-04-08 View on X

Anthropic

Anthropic says Mythos Preview is a general-purpose model and found thousands of high-severity vulnerabilities, including some in every major OS and web browser

Earlier today we announced Claude Mythos Preview, a new general-purpose language model. This model performs strongly across the board …

View original

I was told about the Mythos release, but didn't have access, so have no personal experience to add. Two points from brief: 1) It is not built for IT security, it is just a good enough model that it is good at that too 2) This is the first, not last, model to raise security risks

2026-04-08 View on X

New York Times

Mythos Preview's hacking ability is not a publicity stunt; sources say tech companies privately spoke to Trump officials about the implications for US security

this may shock people — must begin with the two A.I. superpowers, the U.S. and China. It is now urgent that they learn to collaborate to prevent bad actors from gaining access to t...

View original