AI
RESEARCHERS DEVELOP NEW VIRUSES THANKS TO AI
Researchers from the Arc Institute in Palo Alto and Stanford University have developed Evo 1 and Evo 2, two genomic language models capable of analyzing and generating DNA sequences. The concept is similar to ChatGPT or Claude, but applied to living things. Evo has processed thousands of billions of nucleotides, the fundamental building blocks of DNA, to learn the rules that govern genomes. To prevent risks, the researchers deliberately limited their work to a family of bacteriophages, viruses that only infect bacteria. Among the many resulting viruses, sixteen entirely designed by the artificial intelligence model, turned out to be fully functional. They were able to infect their host bacteria and reproduce efficiently. Some even outperformed their natural counterparts.
This ability to quickly generate thousands of variants could significantly shorten research cycles, which were until now limited by traditional experimental trials. But this scientific breakthrough has an obvious downside. Like nuclear energy or some cybersecurity technologies, the value of such a tool depends entirely on the intentions of the person using it. The authors set up strict safeguards by limiting the viruses studied to harmless bacteriophages. Still, there’s no guarantee that all future users will follow the same restrictions…
QUANTUM COMPUTING AND CRYPTOGRAPHY
ALICE & BOB GIVES NOTICE OF A POSSIBLE IMPORTANT FLAW IN POST QUANTUM CRYPTOGRAPHIC ALGORITHMS
We hereby ublish a post by Alice & Bob, a French well known Quantum start up:
A week after an AI model found a weakness in HAWK, a new quantum algorithm claims to break the maths behind most post-quantum cryptography schemes. A preliminary paper just dropped describing a new polynomial-time quantum algorithm against the family of problems that includes lattice cryptography, one of the most famous post-quantum approaches.
Big disclaimer: the paper is not peer reviewed, and the author labeled it preliminary himself. In 2024 a comparable lattice claim was refuted within days. That said, this comes from Daniel Simon, whose 1994 algorithm directly inspired Shor’s, the one that breaks RSA and elliptic curves, most of the encryption today. If it holds (we are already starting to see cracks in the proofs), the stakes are bigger than last week’s. HAWK’s flaw hit one candidate scheme nobody had deployed yet. This claim targets the mathematical assumptions under the whole lattice family, ones NIST has selected to replace RSA and ECC.
Cryptanalysis is active like never before. It is moving on several fronts at once: mathematics, quantum computing, and AI, and we expect more results like this in the coming months. Meanwhile, the quantum computers that would run these algorithms are being built, fast. Replacements have to be trusted before those machines arrive, and post-quantum cryptography cannot afford to be left without options.
Read the full report here
SPACETECH
Brussels and SpaceRISE, a consortium led by SES, Eutelsat, and Hispasat, have finalized an agreement to add 66 satellites in low Earth orbit to the IRIS² program. The project is now moving out of the planning phase and into large-scale deployment. Although this ramp-up will require additional investment, the Commission plans to fund it through a mix of money from the 2028-2034 European budget and contributions from member states and third countries. These 66 new satellites aim to strengthen defense capabilities as well as security and emergency management services. IRIS² will allow authorized administrations in member states, and partner countries like Norway and Iceland, to communicate securely and continuously during sensitive operations (crisis management, protection of critical infrastructure, support for security missions).
