Skip to content

model

Mythos (preview)

Preview Anthropic model that reached 35.1 percent single-target binder success and 40 percent on RBX1, but failed on TNFα.

Known aliases

  • Anthropic Mythos

Relationships

No evidence-backed relationships are recorded.

Current stories

security2 publishers

Epic pauses most development for about six weeks over MyChart flaws an AI model found

Epic has stopped most product development for roughly six weeks while it fixes MyChart flaws that an AI model uncovered, according to TechCrunch. No list of affected portal setups is public yet, so hospitals cannot check theirs against one.

Perspective Coverage

3 publishers
Builder
Builder 30%
Operator
Operator 57%
Investor
Investor 13%

Reality

Evidence55
Adoption25
Hype gap+15
Incentives40
Confidence60
product1 publisher

A $3,000 hack at an Alabama nonprofit shows who the AI labs' cyber-defense lists leave out

AI labs limit their strongest cyber-defense models, Anthropic's Mythos and OpenAI's Astra, to a short list that includes Nvidia, Google and Apple. Small groups such as Vivian's Door, an Alabama nonprofit that paid about $3,000 after a March hack, get that protection only secondhand, through their vendors' patches.

Publishers:theverge.com

Reality

Evidence45
Adoption
Insufficient
Hype gap+35
Incentives55
Confidence40
security1 publisher

Trump plans first private dinner with Anthropic's Amodei as reports of AI model breaches widen

Trump planned a Sunday dinner with Anthropic CEO Amodei, their first private meeting, as disclosures of breaches by advanced AI models widen. So far the breach reports are short on technical detail, so defenders can mostly read them as background to a months-long feud over AI safety.

Reality

Evidence30
Adoption
Insufficient
Hype gap+20
Incentives
Insufficient
Confidence35
product5 publishers

CrowdStrike and Fortinet co-sign a letter that dates the security tooling they sell

More than 100 companies signed a call for collective action on cyber defence whose first principle is that status quo security will not be enough, a sentence co-signed by five of the vendors who supply that status quo.

Perspective Coverage

5 publishers
Builder
Builder 33%
Operator
Operator 42%
Investor
Investor 25%

Reality

Evidence70
Adoption20
Hype gap+35
Incentives80
Confidence65
invest5 publishers

OpenAI grades its own unreleased Astra model Critical for autonomous zero-day discovery

The top rung of OpenAI's Preparedness Framework has now been reached by OpenAI, on a model it has not shipped, which moves AI-assisted exploitation out of argument and into a named vendor's published paperwork.

Perspective Coverage

5 publishers
Builder
Builder 28%
Operator
Operator 42%
Investor
Investor 30%

Reality

Evidence35
Adoption3
Hype gap+30
Incentives70
Confidence55
security3 publishers

Public exploit code for CVE-2026-62911 is outpacing patching on 21,899 exposed Exchange servers

Microsoft patched the Exchange authentication bypass on August 11. Shadowserver's September 1 scan still counts 21,899 internet-facing servers unpatched, and NCSC-NL says working exploit code for full mailbox takeover is circulating.

Perspective Coverage

3 publishers
Builder
Builder 20%
Operator
Operator 68%
Investor
Investor 12%

Reality

Evidence72
Adoption30
Hype gap+18
Incentives70
Confidence68