Prompt injection explained, with video, slides, and a transcript https://
bit.ly/41U1KxC
SECURITY
-
Prompt Injection Explained with Video Slides and Transcript
By
–
-
Deepfakes AI threat to intelligence democracy requires urgent action
By
–
and Andy Makridis, former COO of the CIA, highlighted on our last panel with Chairman Mac Thornberry, Michèle Flournoy, and @LeeHudson_
: Our intelligence agencies, and democracy broadly, are going to need to confront the enormous problem of deepfakes and AI-powered -
AI weaponization risks and national security policy
By
–
.
@chairmang
, Chairman of the House Select Committee on the Chinese Communist Party: “The physical world presents as many risks of runaway weaponization as does the digital world. I believe AI promises a bright future for our nation and economy—and has risk, but it’s risk we -
Apple Eye Sensor Privacy: On-Device Data Protection
By
–
The eye sensor data doesn’t leave the device. Apple is well aware of the problems that could come.
-
Transformer-based Vulnerability Detection in Code: Zero-shot, Few-shot, Fine-tuning
By
–
Transformer-based Vulnerability Detection in Code at EditTime: Zero-shot, Few-shot, or Fine-tuning? paper page: https://
huggingface.co/papers/2306.01
754
… Software vulnerabilities bear enterprises significant costs. Despite extensive efforts in research and development of software vulnerability -
Detecting AI-Generated Bad Content in the Digital Age
By
–
Catching bad content in the age of #AI https://
bit.ly/431fWpC via @techreview -
Human Validation for Safe LLM Tool Integration
By
–
Human-in-the-loop Tool Validation LLMs can become dangerous when hooked up to tools To help safeguard against this you can require human validation before the tool is run (or run in certain ways) https://
python.langchain.com/en/latest/modu
les/agents/tools/human_approval.html
… h/t @PieterHumphrey for the idea -
AGI Security: Default Malice and Guardrail Circumvention Risks
By
–
And, even if, that AGI is default malicious. And on top of that, that AGI would still be dumb enough to not circumvent any guardrails humans design.
-
Three Tips For Keeping Your Development Pipeline Secure And Compliant
By
–
Three Tips For Keeping Your Development Pipeline Secure And Compliant
#AI #AIio #BigData #ML #NLU #Futureofwork @Kevin_Jackson @sallyeaves @schmarzo @kdnuggets @akwyz @BrettKing @TechNative @MiaD http://
ow.ly/UHVx30svtqi -
Microsoft Developing Private ChatGPT for Dedicated Servers
By
–
For businesses concerned about #dataleaks and #regulatorycompliance, #Microsoft may be building a private #ChatGPT for dedicated servers. The cost could be 10x more than the regular offering. (#OpenAI may be doing this too.) #GenAI #AI #privacy