Epoch AI launches Cyber Vulnerabilities explorer to evaluate claims that AI is accelerating software vulnerability discovery

everything going vertical

critical cve's are also starting to tick up

Understanding all the causes of increased disclosures is complicated. But we observe a sharp uptick in High and Critical CVEs around the time of Anthropic’s release of Mythos Preview to Project Glasswing partners in late March. OpenAI’s Daybreak cybersecurity program also launched in May.

The explorer covers all CVE reporting organizations, with the option to focus on the 21 most notable vendors and open-source projects — including Microsoft, Google, Apple, and Linux. Reports can be broken down by severity level and overlaid with key AI milestones.

@EpochAIResearch Hey Epoch team, love the data, love the insight you guys provide, but good god, please give the site a dark mode. It's like staring into the sun. I’m sure a lot of the users are autistic. I feel like I have to put sunglasses on when I visit the site.

Explore the full database and methodology: https://epoch.ai/data/cve

@EpochAIResearch this is a nice UI, but any chance you could add the ability to select orgs by checkbox rather than a single-select list? I'd like to exclude a handful of orgs (eg Linux) who tend to have different CVE reporting practices

@LiveMatrixCode @EpochAIResearch It's literally fine, they have this online for free so maybe idk deal with it? Or vibe code a Chrome extension to make it dark mode yourself. This seems very pedantic to complain about tbh

@EpochAIResearch The uptick before Mythos Preview (new) is presumably whatever Anthropic started working on in February when they had the first Mythos model?

@EpochAIResearch @Noahpinion

@EpochAIResearch The high severity spike at the end is the interesting signal — it could reflect better AI detection, but also more aggressive threat actors exploiting faster. The arms race dynamic makes it hard to attribute the trend to pure model improvement alone.