Lifeboat Foundation

This is a risky bet, given the limitations of the technology. Tech companies have not solved some of the persistent problems with AI language models, such as their propensity to make things up or “hallucinate.” But what concerns me the most is that they are a security and privacy disaster, as I wrote earlier this year. Tech companies are putting this deeply flawed tech in the hands of millions of people and allowing AI models access to sensitive information such as their emails, calendars, and private messages. In doing so, they are making us all vulnerable to scams, phishing, and hacks on a massive scale.

I’ve covered the significant security problems with AI language models before. Now that AI assistants have access to personal information and can simultaneously browse the web, they are particularly prone to a type of attack called indirect prompt injection. It’s ridiculously easy to execute, and there is no known fix.

In an indirect prompt injection attack, a third party “alters a website by adding hidden text that is meant to change the AI’s behavior,” as I wrote in April. “Attackers could use social media or email to direct users to websites with these secret prompts. Once that happens, the AI system could be manipulated to let the attacker try to extract people’s credit card information, for example.” With this new generation of AI models plugged into social media and emails, the opportunities for hackers are endless.

This post is also available in: עברית (Hebrew)

DarkBeam is a digital risk protection firm, which left its interface unprotected and so exposed records with user emails and passwords from previously reported and non-reported data breaches. These leaked logins present cybercriminals with almost limitless attack capabilities.

The CEO of SecurityDiscovery Bob Diachenko first identified the leak, and he states that it contained over 3.8 billion records. DarkBeam has apparently been collecting information to alert its customers in case of a data breach, but this incident will most likely affect more than only DarkBeam users.

In cybersecurity, “threat data feeds” and “threat intelligence” are often used interchangeably. They are, however, quite different. To make matters worse, the term “threat intelligence” has been co-opted and watered down by vendors, making it even more difficult to define the difference between threat data feeds and threat intelligence.

An easy, and accessible, way to tell the difference is to think about weather forecasts. National TV news shows present a forecast for the entire country. You might get some useful information from this, but usually you just get an idea of what the weather is like nationwide. Local weather, however, drills down into the expected conditions for your specific area — not only temperature and weather, but also wind speed, barometric pressure, times for weather changes, and so on. You can use this information to plan out your actions for the next few days.

Using the weather forecast analogy, threat data feeds provide a high-level view of the security landscape. For example, it is useful to know that there is a vulnerability in a specific type of software, but it can be relatively trivial if that software is not in use at your organization. Likewise, knowing which threat groups are active is useful information, but how do you know if they are targeting your sector or organization and what processes and tools they are using?

Today, we are living in the midst of a race to develop a quantum computer, one that could be used for practical applications. This device, built on the principles of quantum mechanics, holds the potential to perform computing tasks far beyond the capabilities of today’s fastest supercomputers. Quantum computers and other quantum-enabled technologies could foster significant advances in areas such as cybersecurity and molecular simulation, impacting and even revolutionizing fields such as online security, drug discovery and material fabrication.

An offshoot of this technological race is building what is known in scientific and engineering circles as a “quantum simulator”—a special type of quantum computer, constructed to solve one equation model for a specific purpose beyond the computing power of a standard computer. For example, in medical research, a quantum simulator could theoretically be built to help scientists simulate a specific, complex molecular interaction for closer study, deepening scientific understanding and speeding up drug development.

But just like building a practical, usable quantum computer, constructing a useful quantum simulator has proven to be a daunting challenge. The idea was first proposed by mathematician Yuri Manin in 1980. Since then, researchers have attempted to employ trapped ions, cold atoms and superconducting qubits to build a quantum simulator capable of real-world applications, but to date, these methods are all still a work in progress.

Dopamine seems to be having a moment in the zeitgeist. You may have read about it in the news, seen viral social media posts about “dopamine hacking,” or listened to podcasts about how to harness what this molecule is doing in your brain to improve your mood and productivity. However, recent neuroscience research suggests that popular strategies to control dopamine are based on an overly narrow view of how it functions.

Dopamine is one of the brain’s neurotransmitters — tiny molecules that act as messengers between neurons. It is known for its role in tracking your reaction to rewards such as food, sex, money, or answering a question correctly. There are many kinds of dopamine neurons located in the uppermost region of the brainstem that manufacture and release dopamine throughout the brain. Whether neuron type affects the function of the dopamine it produces has been an open question.

Recently published research reports a relationship between neuron type and dopamine function, and one type of dopamine neuron has an unexpected function that will likely reshape how scientists, clinicians, and the public understand this neurotransmitter.

Not all wages are bad. If you’re a doctor or nurse you can earn decent. Even a PC tech can earn around $24 an hour or more and cyber security 100k and if you are an engineer you can earn a lot. If you work in a gas station however you won’t earn much. If you are in assembly not much either. It takes skills to earn money. Plus the politicians determine the wages too, it’s not all on big tech. Some tech companies pay more than others of course but knowing AI will increase your wages. They have courses on ChatGPT online now. Even if you run a farm you earn the most money. They’re afraid of progress or I dunno what. Yes we proceed with caution but it’s not like we stop. China won’t nor Russia nor the Middle East etc and even if we’re not in conflict we’ll be left behind.

New book re-examines textile workers’ uprising against the use of technology to erase jobs centuries ago in light of similar problems stemming from AI.

The topic of artificial intelligence’s rising involvement in our digital world and its associated opportunities and challenges have been the main topics of discussion at many security conferences and events in recent times. There is little doubt that humankind is on the verge of an era of exponential technological advancement, and AI is leading the way in the emerging digital world.

For cybersecurity, this tech trend has implications. In simple terms, artificial intelligence acts as a powerful catalyst and enabler for cybersecurity in our connected ecosystem.

The impacts of large language models and AI on cybersecurity range from the good to the bad to the ugly. Here’s what to watch out for, and how to prepare.

Sandman, a new cyber threat actor, is targeting telecom providers across continents.