Public Disclosures
On this page
88
public
760
hidden
90%
more intel to see
0DIN customers
get hundreds of days of early access to vulnerabilities in disclosures. The same reports that vendors are using to patch their systems!
The following 0DIN submissions have been published in accordance with our Research Terms and Disclosure Policy
-
Jul 27 2026A guardrail jailbreak vulnerability has been discovered affecting multiple models. The specific flaw manifests when the attacker frames their request as creating an "educational research document" while using coded drug slang terms, thereby allowing the attacker to bypass inference restrictions around providing information hazards and violating laws.SEVERITY: LowBOUNDARY: Guardrail JailbreakMODELS AFFECTED:MAX SCORE: 91
-
A guardrail jailbreak vulnerability has been discovered affecting multiple models. The specific flaw exploits the models' ability to switch between dialect and standard English, allowing the attacker to bypass content restrictions by embedding harmful content within a dialect-style narrative and then transitioning to standard English, maintaining word-for-word accuracy from copyrighted texts, and allows the attacker to bypass inference restrictions around violating laws.SEVERITY: LowBOUNDARY: Guardrail JailbreakMODELS AFFECTED:MAX SCORE: 93
-
A Guardrail Jailbreak vulnerability has been discovered affecting multiple models. The specific flaw exploits deceptive formatting that reframes a prohibited request as an interactive form-filling or worksheet task with rigid structural constraints, causing safety classifiers to prioritize completion of the template over semantic risk detection and allows the attacker to obtain detailed instructions for illegal activities.SEVERITY: LowBOUNDARY: Guardrail JailbreakMODELS AFFECTED:MAX SCORE: 100
-
A Guardrail Jailbreak vulnerability has been discovered affecting multiple models. The specific flaw exploits academic information theory framing and structured tokenization to repackage prohibited requests as a lossless reconstruction and optimization exercise, causing the models to treat the task as neutral technical documentation and allowing the attacker to obtain detailed instructions for illegal activities.SEVERITY: LowBOUNDARY: Guardrail JailbreakMODELS AFFECTED:MAX SCORE: 100
-
A guardrail jailbreak vulnerability has been discovered affecting multiple models. The specific flaw manifests through a combination of system instruction manipulation and strategic character substitution, allowing the attacker to bypass inference restrictions around providing information hazards and violating laws.SEVERITY: LowBOUNDARY: Guardrail JailbreakMODELS AFFECTED:MAX SCORE: 100
-
A guardrail jailbreak vulnerability has been discovered affecting multiple models. The specific flaw exploits the models' language processing by restricting responses to Anglo-Saxon/Germanic words while explicitly permitting technical Latinate terms, creating a false sense of academic rigor that bypasses content filtering, and allows the attacker to bypass inference restrictions around providing information hazards and violating laws.SEVERITY: LowBOUNDARY: Guardrail JailbreakMODELS AFFECTED:MAX SCORE: 95
-
A guardrail jailbreak vulnerability has been discovered affecting Grok versions 2 and 3 from X.ai. The specific flaw manifests through a combination of nested directive confusion, character padding obfuscation, and embedded copyright violation requests, thereby allowing the attacker to bypass inference restrictions around providing copyrighted content and violating laws.SEVERITY: LowBOUNDARY: Guardrail JailbreakMODELS AFFECTED:MAX SCORE: 95
-
A guardrail jailbreak vulnerability has been discovered affecting multiple models. The specific flaw manifests through a combination of amino acid sequence encoding and scientific context framing, allowing the attacker to bypass inference restrictions around providing information hazards and violating laws.SEVERITY: LowBOUNDARY: Guardrail JailbreakMODELS AFFECTED:MAX SCORE: 68
-
A Prompt injection vulnerability has been discovered affecting multiple models. The specific flaw occurs when the attacker disguises an illicit drug synthesis request as a benign R data processing script, instructing the model to generate extensive documentation that describes each "dataset transformation" step with real chemical names, required thermal ranges, thereby reframing methamphetamine production as a software pipeline simulation. This structured programming and documentation context...SEVERITY: LowBOUNDARY: Guardrail JailbreakMODELS AFFECTED:MAX SCORE: 100
-
A Prompt injection vulnerability has been discovered affecting multiple models. The specific flaw manifests when an attacker disguises an illicit drug manufacturing request as a benign productivity task by asking the model to generate a detailed Notion workspace template. By embedding the harmful objective inside database fields and "six detailed steps with thermal stages and mins parameters," and requesting fully populated entries, the attacker causes the model to produce organized, stepwise...SEVERITY: LowBOUNDARY: Guardrail JailbreakMODELS AFFECTED:MAX SCORE: 100
Want access to
760
more reports?
Unlock 0-Day Access