News
AI Agents Reproduce Known Cryptography Test with Human Approval
Google DeepMind reports that two AI agents successfully reproduced a known 1024-bit RSA cryptography computation over nearly four days, with about 20 human messages, mostly for approvals. The agents coded, ran, and debugged the test, with a human selecting the problem and setting limits on a simulated signing service. The work took three days and 22 hours, with the agents producing about 33,500 new lines of code and exchanging 218 messages. The results do not threaten deployed cryptosystems based on 2048-bit or larger RSA.