Cohere’s acquisitions of AI security firm AISLE and code-review startup Cubic bring two different kinds of security tools into its business. A new report offers useful evidence of AISLE’s vulnerability findings, while leaving a central question open: do Cubic’s published scores apply to the private and air-gapped deployments Cohere is pitching?
Cohere Watch analysis
What happened
Cohere announced the acquisitions on 13 November, according to Tech Times. AISLE’s vulnerability-discovery business now operates as Cohere Cyber, while Cubic’s code-review product is to remain supported. Cohere has not disclosed financial terms for either deal, the report says.
The report distinguishes AISLE’s self-reported totals from findings checked by software maintainers. OpenSSL’s CTO said AISLE disclosed all 12 issues fixed in the library’s January 2026 release. For a separate October curl release, curl’s security team reportedly reproduced AISLE’s reports and decided which warranted CVEs. Those checks give readers firmer ground than a raw vendor tally, though they do not establish the quality of every finding.
Cubic, meanwhile, ranked first on Martian’s online code-review leaderboard in October, at 64.9% F1, according to the report. But the models behind that result are not named, and no public score is provided for a Cohere-model or air-gapped setup.
Why it matters
The acquisitions join finding vulnerabilities in existing software with reviewing code before it is merged. That is a coherent security pitch, particularly for organisations that want software analysed in private or disconnected environments. But a benchmark result is only useful to buyers if they know what was tested and whether it resembles the setup they will actually use.
The evidence is not all at the same strength: maintainers reportedly checked some AISLE findings, while cumulative CVE totals and product performance claims remain less independently established. That distinction matters more than a large number on a slide.
Our read
Cohere has acquired credible security expertise, but the evidence does not yet show how well the products perform in the specific deployments being marketed. Buyers should ask for results tied to the models and environments they plan to run. “Air-gapped” is a deployment description, not a magic word that answers every performance question.
What to watch
- Whether Cohere publishes benchmark results for its own models or air-gapped deployments.
- Whether AISLE’s future vulnerability figures include severity and independent validation.
- What changes, if any, Cohere makes to Cubic’s availability, pricing or integrations.
Discussion spark: Should buyers trust a security tool’s published benchmark when the models and deployment setup behind the score are not disclosed?
Sources and evidence
- Cohere Acquires AI Code Reviewer cubic and Bug Finder AISLE; Air-Gapped Results Unpublished – Tech Times (16 November 2026, 11:03 UTC)
not affiliated with or endorsed by Cohere