Tag: large language model
AI gives the game away
The latest threat for companies using large language (LLM) AI software to replace human staff is the software’s innate gullibility. LLM software can be likened to some cowardly bank clerk in an old Western hold-up who not only willingly opens a back door for the bad guys but also willingly tells them the combination of the safe. The methods for persuading LLMs into naively disclosing the keys to the corporate kingdom are known as ‘LLM Jailbreak’ techniques. Palo Alto Networks Unit 42 researchers have named one such LLM Jailbreak, “Bad Likert Judge”.
Security minefield ahead for GenAI users
In an exclusive interview with Cyber Intelligence, Gadi Bashvitz, CEO of cybersecurity testing firm, Bright Security warns of the security challenges facing organizations in the wake of widespread adoption of GenAI. Cyber Intelligence: Are there any specific dangers of which companies using GenAI to generate new code should be particularly aware? Gadi Bashvitz: There are multiple considerations here. On one hand, any solution developed leveraging LLMs is prone to LLM-specific vulnerabilities such as Insecure Output Handling and Broken Access Control and it is critical to make sure organizations are aware and can detect such vulnerabilities before releasing LLM-based solutions.
Apple enters GenAI marketplace
Apple has joined Google and Microsoft in launching its own generative artificial intelligence (AI) offering, OpenELM. Apple claims that OpenELM, “a state-of-the-art open language model,” will offer users more accurate and less misleading results than its widely criticized competitors. “OpenELM uses a layer-wise scaling strategy to efficiently allocate parameters within each layer of the transformer model, leading to enhanced accuracy,” says Apple. Apple claims that OpenELM exhibits a 2.36 percent improvement in accuracy compared to its initial predecessor OLMo, while requiring half as many pre-training tokens. So far, Apple has delayed offering modern AI capabilities on its devices, but it is expected that the next version of its operating systems will need to include some unique AI features. The launch of iOS 18 is scheduled for June 10.
Cybersecurity Executive Admits to Hospital Hacking for Business Gain – November 20th
In a startling revelation, Vikas Singla, the former COO of cybersecurity firm Securolytics, confessed to hacking two Georgia hospitals in June 2021 to enhance the company’s profile. Singla disrupted services at Gwinnett Medical Center hospitals, stealing patient data and publicizing the breach on Twitter. Facing 17 counts of computer damage and one count of information theft, Vikas Singla agreed to pay over $817,000 in restitution. Due to health issues, prosecutors recommended 57 months of probation, raising concerns about cyber threats jeopardizing public safety and healthcare data.