The Blacklynx Brief - Vol.5 - Evil AI Chatbots

Welcome to the Blacklynx Brief for the week of September 3rd. We’re in the last throes of summer - the fall is lurking right around the corner and in the world of AI and breaches the news keeps coming out relentlessly.

We’re a brand new publication and we need YOU (yes YOU!) to help us grow.

There’s a lot of work that goes into these newsletters. I mean, hours of research and writing. To be honest I was a bit blindsided by the work this takes to put together but I’m committed to keep doing this.

The only thing we ask in return is to PLEASE share this in some way in your network. Either by using one of the social media buttons in the email or by mentioning it on LinkedIn. Anything ! 😃 

It’s only a second for you and it would help us tremendously.

Currently we’re at 93 subscribers (I screwed up this week but that is another story)

We’re kicking of the newsletter with all the latest from the cyber world. Strap in !

Cybersecurity News

The cyber news was a little bit calmer this week - but as if to validate the existence of this publication the one that caught our eye was the first on this list. Who would’ve thought the bad guys have started using AI in their nefarious activities (sly smile) ?

AI Large Language Model Prompt Injection Threats

The NCSC - the UK’s National Cyber Security Centre has published a blog about the threat prompt injection attacks pose to Large Language Models (LLM’s) like ChatGPT. This involves creating input that “tricks” these language models in doing things that are beyond the safety boundaries developers put around the models. This might be malicious behavior like the generation of offensive content, exposing data or other evil acts.

Targeting Ukranian Soldier Phones

Intelligence reports from the so-called “Five Eyes Countries” —covering Australia, Canada, New Zealand, the UK, and the US—reports that Sandworm hackers are going after Ukrainian soldiers' phones.

Sandworm is an infamous Russian hacker group that is well known for taking down part of the Ukranian powergrid in December 2015. If you want to dig deeper into the murky world of Russian hacker groups I recommend reading Andy Greenberg’s “Sandworm”

They're using malware called “Infamous Chisel-A”. This nasty code does a lot: steals files, opens a backdoor, and can even take over network traffic and file transfers. The war in Ukraine has a covert cyber component that is by nature difficult to uncover. But it’s not only missiles and bullets.

If you’re inclined to delve into deep technical details: go nuts

High Security Military Fencing Company in UK gets breached

Zaun is a company that manufactures high-security perimeter fences for the UK’s military sites. Rabbit hole alert : Researching this company I spent an hour looking at all kinds of fences. It’s a world that I didn’t know existed 😀 

Anyway - they got breached by Lockbit Ransomware hackers and 10 GB was exfiltrated from the network. In a statement on their website the breach was likely caused by a PC running Windows 7.

The irony. You build high-security hardware to keep people out only but leave gaping holes in your software.

A stark reminder to get rid of everything that is not Windows 10 or above.

Freecycle Breach : More than 7 Million (!) Users Impacted

Non-profit organization Freecycle Network (Freecycle.org) hast confirmed it has suffered a data breach.

More than 7 million user’s usernames, user IDs, email addresses and passwords were stolen.

The company issued a statement and has urged all members to change their password. That’s a rather lacklustre approach if you ask me. Especially if you read the statement it feels like a simple sweep under the rug.

On to the world of AI …

AI News

Tesla's Advances in Autonomous Driving

Elon Musk demonstrated Tesla's fully self-driving V12 beta software. The software claims to be FULL Artificial Intelligence. It operates without an internet connection and navigates complex traffic scenarios. It makes decisions on the spot based on the feedback of millions of Tesla drivers.

Interestingly this is not always compliant with traffic regulations. For example only 0,5% of drivers come to a complete stop when driving up to a “STOP”-sign. So the AI also doesn’t come to a complete stop. Which is a problem …

Tesla also unveiled a $300 million supercomputer for their AI training. This thing contains about 10.000 Nvidia H100 GPU’s. It’s being described as “one of the most powerful machines in the world”.

OpenAI and Copyright Issues

OpenAI is facing legal challenges for using copyrighted material to train their language models. They argue that their usage falls under "fair use," a defense that could set a precedent for AI training methodologies. The U.S. Copyright Office is also seeking public opinion on AI and copyright, which could lead to new regulations affecting the AI industry.

ChatGPT Enterprise Version

Few weeks ago we talked about OpenAI being in financial trouble. In response to that OpenAI introduced an enterprise version of ChatGPT, offering enhanced features and data analysis capabilities. This could attract businesses looking for advanced AI solutions, thereby increasing OpenAI's market share and revenue.

AI Developments in China

China has launched new AI models, including Qwen-VL-Chat and Ernie bot. These models offer functionalities like image reading and equation solving. The Chinese market's rapid advancements in AI could pose competition to Western companies, affecting global market dynamics.

Google's AI Initiatives

Google announced several AI-related features- like Duet AI for Google Workspace, including digital watermarking and a partnership with Nvidia. They also expanded their AI services to India and Japan. These moves could strengthen Google's position in the AI market. Very interesting !

AI in Food and Delivery Services

Samsung's AI-powered recipe app and AI initiatives from DoorDash and Uber Eats indicate a growing trend of AI applications in the food and delivery sector. These could offer personalized customer experiences, thereby increasing user engagement and sales.

Substack's AI-Powered Transcription

Substack has integrated AI to transcribe podcasts into written content. This feature could attract more content creators to the platform, increasing its user base and potential for ad revenue.
That’s it for us this week.

I’ll leave you with an AI-generated image from the latest Midjourney build

May I ask you - if you found ANY value in these words, may we kindly ask it for you to SHARE it with friends, family or colleagues who might find it intriguing as well?

Whether it’s forwarding the email or sharing it on social media, your endorsement means the world to us.

It’s a second for you - it’s life and death for us :)

Warm regards, and see you next week !