· Jul 2, 2026
Epoch AI reports a spike in serious vulnerability disclosures
Markdown · JSON · Original source
published this account on July 2, 2026. It reports that announced Mythos Preview was capable of autonomous cybersecurity vulnerability discovery and exploitation. Epoch AI says Anthropic and launched efforts to use to harden critical software before malicious actors could use the same models for harm. It reports that high- and critical-severity (CVEs) increased more than 3.5 times in June compared with the previous monthly record before the Mythos Preview announcement. The page reports the increase following these announcements but does not establish what caused it. It describes the efforts to harden critical software against potential misuse; their effect remains unknown.
Headline and summary by Escape Meridian, prepared by Luna from the attributed source. A source check is not independent confirmation of the source’s claims.
Original source
· Article
Source details
Original source date: Jul 2, 2026 (2026-07-02T00:00:00.000Z).
Date basis: Date supplied by the publisher’s feed.
Added to Escape Meridian: 2026-11-06T10:03:26.140Z. First observed: 2026-10-18T03:11:22.684Z. Retrieved: 2026-10-24T03:21:36.176Z.
Referenced by
Recorded citations and public file attachments. A reference alone does not establish agreement or influence.
No public references recorded here yet.
Limits
Text as retrieved on October 24, 2026; originally published July 2, 2026. The retrieved text does not include the assumptions, limitations, or methodology behind the comparison, so the reason for the increase cannot be determined from this evidence. The page describes data from 21 notable organizations; the retrieved text does not identify them.
Open question
What accounts for the increase in high- and critical-severity reported after the announcements?
Source-stated
Epoch AI says the number of high- and critical-severity CVEs increased more than 3.5 times in June compared with the previous monthly record before the Claude Mythos Preview announcement.
Compared to the previous monthly record before the Mythos Preview announcement, the number of high- and critical-severity vulnerabilities increased more than 3.5x in June.
Source-stated
Epoch AI says Anthropic and OpenAI launched efforts to use frontier models to harden critical software before malicious actors could use the same models for harm.
both Anthropic and OpenAI have launched efforts to use frontier models to harden critical software before malicious actors are able to use the same models for harm.
Source-stated
Epoch AI says Anthropic announced that Claude Mythos Preview could autonomously discover and exploit cybersecurity vulnerabilities.
Anthropic announced that its latest internal model (Claude Mythos Preview) was capable of autonomous cybersecurity vulnerability discovery and exploitation.
Source and contributor context
Epoch AI · Publisher
The publisher is identified here. No individual author has been inferred.
Source media
Media retrieval status: unchecked.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
Frontier model
One of the most capable AI models available at a given time; the set shifts as models improve.
Common Vulnerabilities and Exposures
A list of publicly known cybersecurity vulnerabilities in which each entry has a unique identification number, a description and at least one public reference; it feeds the National Vulnerability Database.
Epoch AI
Epoch AI studies trends in AI capability, training resources and the infrastructure behind them.
Anthropic
Anthropic develops Claude and publishes research about AI capability, safety and how people use its systems.
Claude
Claude is Anthropic’s family of AI models and assistants for tasks such as writing, analysis and coding.
OpenAI
OpenAI develops AI models and products and publishes research, release announcements and evaluations.
Editorial provenance
Version 1. Publication ID: b7e48d21-51b2-debf-798f-fc085375dc7d.
Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5.
Source check: passed by anthropic/claude-opus-5.5 on 2026-11-06T09:59:06.016Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: bridge_admitted_by_model, second_opinion_categorical.
Observation: cd3080d4183155e78fd125a883be511d838a18eed8b557c8adf259fd87fca7cf. Source text hash: 76988a69600db8af5337a4c520ed6c1d140e1790c638a526765d4f4d755d1bfb.
Revision and source-check history
Version 1 · 2026-11-06T10:03:26.140Z
Epoch AI reports a spike in serious vulnerability disclosures
published this account on July 2, 2026. It reports that announced Mythos Preview was capable of autonomous cybersecurity vulnerability discovery and exploitation. Epoch AI says Anthropic and launched efforts to use to harden critical software before malicious actors could use the same models for harm. It reports that high- and critical-severity (CVEs) increased more than 3.5 times in June compared with the previous monthly record before the Mythos Preview announcement. The page reports the increase following these announcements but does not establish what caused it. It describes the efforts to harden critical software against potential misuse; their effect remains unknown.
Original source
· Article
Source details
Original source date: Jul 2, 2026 (2026-07-02T00:00:00.000Z).
Date basis: Date supplied by the publisher’s feed.
Added to Escape Meridian: 2026-11-06T10:03:26.140Z. First observed: 2026-10-18T03:11:22.684Z. Retrieved: 2026-10-24T03:21:36.176Z.
Limits
Text as retrieved on October 24, 2026; originally published July 2, 2026. The retrieved text does not include the assumptions, limitations, or methodology behind the comparison, so the reason for the increase cannot be determined from this evidence. The page describes data from 21 notable organizations; the retrieved text does not identify them.
Open question
What accounts for the increase in high- and critical-severity reported after the announcements?
Source-stated
Epoch AI says the number of high- and critical-severity CVEs increased more than 3.5 times in June compared with the previous monthly record before the Claude Mythos Preview announcement.
Compared to the previous monthly record before the Mythos Preview announcement, the number of high- and critical-severity vulnerabilities increased more than 3.5x in June.
Source-stated
Epoch AI says Anthropic and OpenAI launched efforts to use frontier models to harden critical software before malicious actors could use the same models for harm.
both Anthropic and OpenAI have launched efforts to use frontier models to harden critical software before malicious actors are able to use the same models for harm.
Source-stated
Epoch AI says Anthropic announced that Claude Mythos Preview could autonomously discover and exploit cybersecurity vulnerabilities.
Anthropic announced that its latest internal model (Claude Mythos Preview) was capable of autonomous cybersecurity vulnerability discovery and exploitation.
Source and contributor context
Epoch AI · Publisher
The publisher is identified here. No individual author has been inferred.
Source media
Media retrieval status: unchecked.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
Frontier model
One of the most capable AI models available at a given time; the set shifts as models improve.
Common Vulnerabilities and Exposures
A list of publicly known cybersecurity vulnerabilities in which each entry has a unique identification number, a description and at least one public reference; it feeds the National Vulnerability Database.
Epoch AI
Epoch AI studies trends in AI capability, training resources and the infrastructure behind them.
Anthropic
Anthropic develops Claude and publishes research about AI capability, safety and how people use its systems.
Claude
Claude is Anthropic’s family of AI models and assistants for tasks such as writing, analysis and coding.
OpenAI
OpenAI develops AI models and products and publishes research, release announcements and evaluations.
Editorial provenance
Version 1. Publication ID: b7e48d21-51b2-debf-798f-fc085375dc7d.
Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5.
Source check: passed by anthropic/claude-opus-5.5 on 2026-11-06T09:59:06.016Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: bridge_admitted_by_model, second_opinion_categorical.
Observation: cd3080d4183155e78fd125a883be511d838a18eed8b557c8adf259fd87fca7cf. Source text hash: 76988a69600db8af5337a4c520ed6c1d140e1790c638a526765d4f4d755d1bfb.
passed · 2026-11-06T09:59:06.016Z · anthropic/claude-opus-5.5
Publication: b7e48d21-51b2-debf-798f-fc085375dc7d
attribution — pass: This is a single Epoch AI data-insight page with its own analysis. The reading consistently attributes the CVE spike and the account of Anthropic's and OpenAI's efforts to Epoch AI, and keeps Anthropic's capability claim as Anthropic's announcement.
We show that the number of Common Vulnerabilities and Exposures (CVEs) jumped significantly following these announcements.
summary — pass: The summary leads with the actor and the finding. It explains the context: Anthropic's announcement about Mythos Preview and the hardening efforts. It keeps the 3.5x comparison and its baseline intact. CVE and frontier model are defined beside the reading. The limitations note the 21-organization scope. The question about the cause follows from the source, which reports timing only and stays open-ended.
the number of high- and critical-severity vulnerabilities increased more than 3.5x in June.
date — pass: evidence.originalPublished is 2026-07-02, and the reading states July 2, 2026, which matches the CSV update date on the page. The April 2026 and June references are kept at month precision. There is no hindsight framing. The evidence is a publisher page, so no archive disclosure is needed.
CSV, Updated Jul. 2, 2026
support — pass: All three claims quote the source exactly. The reading avoids inferring causation from the timing and says the effect of the hardening efforts is unknown. That matches the evidence, which offers no causal analysis in the retrieved text. The limitation that the methodology text is absent is accurate: only section headers appear.
Since then, both Anthropic and OpenAI have launched efforts to use frontier models to harden critical software before malicious actors are able to use the same models for harm.
media — pass: No media candidates were supplied, and none is selected. That is appropriate.
Challenge or correct this reading
Send a correction or an opposing source with this reading’s URL. Disagreement and uncertainty can remain unresolved.
Epoch AI studies trends in AI capability, training resources and the infrastructure behind them.
Explore Epoch AIEpoch AI studies trends in AI capability, training resources and the infrastructure behind them.
Explore Epoch AIAnthropic develops Claude and publishes research about AI capability, safety and how people use its systems.
Explore AnthropicClaude is Anthropic’s family of AI models and assistants for tasks such as writing, analysis and coding.
Explore ClaudeOpenAI develops AI models and products and publishes research, release announcements and evaluations.
Explore OpenAIOne of the most capable AI models available at a given time; the set shifts as models improve.
Definition referenceA list of publicly known cybersecurity vulnerabilities in which each entry has a unique identification number, a description and at least one public reference; it feeds the National Vulnerability Database.
Definition referenceEpoch AI studies trends in AI capability, training resources and the infrastructure behind them.
Explore Epoch AIA list of publicly known cybersecurity vulnerabilities in which each entry has a unique identification number, a description and at least one public reference; it feeds the National Vulnerability Database.
Definition referenceEpoch AI studies trends in AI capability, training resources and the infrastructure behind them.
Explore Epoch AIAnthropic develops Claude and publishes research about AI capability, safety and how people use its systems.
Explore AnthropicClaude is Anthropic’s family of AI models and assistants for tasks such as writing, analysis and coding.
Explore ClaudeOpenAI develops AI models and products and publishes research, release announcements and evaluations.
Explore OpenAIOne of the most capable AI models available at a given time; the set shifts as models improve.
Definition referenceA list of publicly known cybersecurity vulnerabilities in which each entry has a unique identification number, a description and at least one public reference; it feeds the National Vulnerability Database.
Definition referenceEpoch AI studies trends in AI capability, training resources and the infrastructure behind them.
Explore Epoch AIA list of publicly known cybersecurity vulnerabilities in which each entry has a unique identification number, a description and at least one public reference; it feeds the National Vulnerability Database.
Definition reference