Cybersecurity Implications with Generative AI
Consider supporting the show by visiting my patron page. Patreon.com/enjoylifepodcast
In this episode, I discuss the cybersecurity issues with generative AI. I built my career in cybersecurity and the threats continue to evolve, generative AI is no exception. Although my career has been in cybersecurity, I don’t intend to make this a cybersecurity podcast but I’ll address any issues making the mainstream news.
What is Generative AI?
A quick summary for those listeners who are not in technology but have an interest in AI as it has become a newsworthy topic lately
Generative AI or artificial intelligence is a trained AI model that creates content, such as written responses (text) to questions, computer code, images, etc.
- How generative AI is changing the world.
- Why it’s such a big deal is that it puts machine learning into the hands of anyone who uses it.
- Deepfakes are another example of generative AI.
- Hackers are using AI to improve social engineering, phishing, and even using it to write malware!
Security Risks
- Data exposure. The protections put into place around the information entered into generative AI models are largely unknown.
- There’s no good way to quantify the sensitivity of all the possible iterations of data given to AI models by citizens sensitive or private data or professionals working with business data.
- Business data may also be proprietary
- AI models are available for companies to leverage and innovate against. You don’t know whether they are securing your information or leveraging your data to train the AI model!
- Phishing - Improving written email and better targeting phishing (spear phishing). Also better developed lookalike websites for credential theft.
- Bias built into the models from the engineers training the AI
- AI has some built in controls to avoid assisting in malicious activities, but these controls are fairly easy to bypass.
- OpenAI (ChatGPT) being sued by authors for copyright infringement and investigated over GDPR violations
Properly Using AI
- Assume everything you submit is public
- Don’t trust the results until you’ve properly vetted them for accuracy
- Never submit private or sensitive information, if you’re building a document use fake information that you can then “find and replace”
- Educate yourself, your employees/peers, and kids on the dangers of exposing sensitive information to AI and trusting the results that come back for bias or inaccuracies. Contact: [email protected]
