OpenAI's Secret Breakthrough

Good morning. Another friday. Another week filled to the brim with news.

It’s the Blacklynx Brief not the Blacklynx Rumor Mill so I will resist the temptation to dive too deep into all the rumors that are swirling around about the board switch at OpenAI and their “secret breakthrough””

Time will tell what is fact and what proves to be fiction.

But I do recommend, if you have any technical inclination, watching the video from someone in a Star Trek costume explaining what is potentially going on (see below)

[AI]
OpenAI’s Secret Breakthrough

According to a report in Reuters - OpenAI had a secret breakthrough called Q* (Q-star) that precipitated the firing of Sam Altman.

Before Altman’s firing the researchers sent the board an official “letter of warning” that the discovery could “threaten humanity”. (By the way I’m not making any of this up).

This new model has shown the internal capabilities of teaching itself simple math. That is from the Reuters article.

However, there was an additional leaked document someone posted on the very doubtful 4chan website. This document was analyzed by experts and excellently so in this video (don’t mind the fact that the presenter is sitting there in a Star Trek costume).

The gist of it is that this Q* algorithm apparently, after starting on basic math - was able to break a 192-bit cryptographic cypher.

If this is true this newsletter is going to become even more interesting. Especially the cybersecurity part.

[Cyber]
Cyber Insurance Takes Hold in the EU

A recent report from ENISA, the EU's cybersecurity agency, reveals a growing trend in cyber insurance among Europe's essential service providers and digital companies. In 2022, 42% of these crucial entities have opted for cyber insurance, a significant increase from the 30% in 2021 and close to the 43% in 2020. This marks a change from the past, where only a few member states showed significant uptake in cyber insurance.

The data, gathered from over a thousand organizations across the EU, reflects an increased commitment to meeting the EU's Network and Information Systems (NIS) directive. This directive mandates that large organizations enhance their network security and report major cybersecurity incidents.

Interestingly, while cybersecurity spending has slightly increased, with the average organization dedicating €5.1 million of their €83.6 million IT budget to it, there's a notable stability in how companies are staffing their cybersecurity teams. Around half of the organizations do not plan to increase their cybersecurity staff in the next two years, suggesting a newfound balance in both budgeting and staffing.

However, with the upcoming transition to the new NIS2 directive, a surge in cybersecurity investment is anticipated across the EU, signaling a re-emphasized focus on digital safety in a rapidly evolving technological landscape.

[AI]
This Week in AI

• Text to video is making strides and AI startup Pika is on the forefront of this new family in generative AI. They just announced a 55M$ funding round. The potential and implications of this are enormous. Watch it in action here

• AI is being tested in the medical field and GPT-4 proves to be amazing at radiology. A research paper from Microsoft demonstrates strong performance on radiology tasks like summarization and triage, achieving state-of-the-art results.

• According to the New York Times, US officials are scrutinizing the UAE's AI giant G42 for its ties with Chinese firms, citing potential risks in data and technology transfer. G42, known for its recent partnership with OpenAI and collaborations with tech majors like Microsoft and Dell, is under the microscope for its connections with companies like Huawei, raising concerns over the transfer of advanced U.S. technology to China's military. The company's CEO, Peng Xiao, previously managed ToTok, which is flagged by U.S. intelligence as a Chinese-supported espionage tool. (I guess this is cybernews as well ;))

• The US, UK and seventeen other countries signed an agreement to make AI “secure by design” in what i guess is another cybersecurity story.

• The Pentagon is going all out on the development of lethal AI-driven drones and autonomous defense systems. “To keep pace with China”. Project “Replicator” should finish in 2026 and the goal is to have fully autonomous AI driven devices running perimeters and doing missions. What can go wrong?

• While many people are screaming to make AI development safer, NVIDIA’s Jensen Huang calls for an acceleration in development. He just wants to sell more circuitboards I guess

[Cyber]
This Week in Cyber

• In a dramatic twist, more than 80 UK law firms are scrambling as their cloud host, CTS, fell victim to a CitrixBleed attack. Yep, the notorious LockBit ransomware gang strikes again. Since last Wednesday, real estate transactions have hit a virtual wall, leaving firms and their clients in limbo.

• Not a good time for Bahrain’s Gulf Air, as they've just announced a breach that might have exposed customer data. Luckily, their planes are still flying high, as critical systems and flight operations weren’t affected

• In the US, Ardent Health's hospitals across six states are in crisis mode, thanks to a Thanksgiving Day ransomware attack. Ambulances rerouted, patients redirected – it's a real medical drama, minus the TV cameras. Ardent's quick shutdown of its IT network is like applying a band-aid to a gaping wound.

• Aliquippa, Pennsylvania, now has a new water crisis on its hands. An Iranian group, Cyber Av3ngers, claimed responsibility for hijacking the town's water booster stations. Unitronics, the manufacturer, is probably having a major headache right now. CBS is all over this story, and so are worried residents.

• Here’s a new one: Google Drive users have suddenly found their files, dating back to May 2023, missing in action. No trash folder recovery, no web version backups. Just poof, gone! Google's scratching their heads, and so are we. Android Police has the scoop, but no solution yet.

• Lastly, Meta’s in hot water in the EU, thanks to a complaint from privacy watchdog noyb. Their new scheme? Charging EU folks €9.99/month to dodge online tracking. But noyb's calling foul, saying privacy isn't a luxury item. And about those "psychopathic tech bros" – let's just say I might've added that for flavor!

[Closing Thoughts]
What We Learned

We learned that the ruthlessness of hackers knows no bounds : water installations, hospitals , Google Drive , … these people have no human decency whatsoever.

We also still haven’t learned the official version of what has transpired at OpenAI two weeks ago.

See you next week with hopefully more answers and less data breaches