#AIvulnerability

How to jailbreak ChatGPT and trick the AI into writing exploit code using hex encoding

by

October 29, 2024 at 06:36PM OpenAI’s GPT-4o can be manipulated into generating exploit code by encoding malicious instructions in hexadecimal, bypassing its safety features. Researcher Marco Figueroa highlights this vulnerability on Mozilla’s 0Din platform, emphasizing the need for improved AI security measures and detection mechanisms for encoded content to prevent such exploitations. ### Meeting Takeaways … Read more

Slack Patches AI Bug That Let Attackers Steal Data From Private Channels

by

August 22, 2024 at 11:47AM Salesforce’s Slack AI has patched a flaw identified by security firm PromptArmor, which could have allowed attackers to steal data from private Slack channels or engage in secondary phishing within the platform. The flaw is related to the use of a language model that did not recognize malicious instructions, enabling … Read more

Experts Find Flaw in Replicate AI Service Exposing Customers’ Models and Data

by

May 25, 2024 at 06:18AM A critical security flaw in AI-as-a-service provider Replicate allowed unauthorized access to proprietary AI models and sensitive information due to a vulnerability in its containerization process. The flaw was responsibly disclosed and addressed, and there is no evidence of exploitation. However, it highlights the potential risks of malicious models in … Read more

‘Conversation Overflow’ Cyberattacks Bypass AI Security to Target Execs

by

March 19, 2024 at 08:06AM AI email security controls are being bypassed by credential-stealing emails that hide malicious payloads within harmless-looking emails. This poses a major threat to enterprise networks. After reviewing the meeting notes, the key takeaways are: 1. Credential-stealing emails are bypassing AI’s “known good” email security controls by disguising malicious payloads in … Read more