Bill Gates issues warning on 'bad people' exploiting a lack of AI safeguards to commit cyberattacks

Microsoft co-founder Bill Gates warned Sunday that artificial intelligence could fuel a wave of cyberattacks over the next year.

Bill Gates addressed the Munich Security Conference on February 18 2024, warning that AI‑driven cyberattacks could climb by roughly 40 % over the next year if safeguards lag. He pointed to Microsoft’s threat‑intelligence data showing a 28 % rise in AI‑generated phishing emails in Q4 2023 compared with the prior quarter. Gates noted that unrestricted open‑source large language models are being downloaded more than a million times each month, giving malicious actors easy access to powerful tools.

The Federal Trade Commission recorded a 22 % increase in consumer complaints about AI‑generated scams during January and February 2024. In response, insurers are adjusting premiums for identity‑theft protection policies, projecting an average monthly rise of about $5 per policy by Q3 2024. If you subscribe to a typical $9.99‑per‑month monitoring service, your bill could reach roughly $15 per month later this year.

After the 2021 Colonial Pipeline ransomware incident, the US issued Executive Order 14028 mandating software bills of materials for federal contractors, yet no comparable rule exists for AI model distribution. Microsoft’s internal AI ethics board recommended mandatory provenance tags for all large language model releases in 2022, a proposal that remained advisory and was not adopted company‑wide. As of February 2024 the EU AI Act awaits a final plenary vote, leaving the United States without federal AI‑specific cybersecurity regulations.

The Senate Commerce Committee is slated to vote on March 12 2024 on a bill that would require AI model providers to implement usage‑logging and anomaly‑detection controls. Microsoft plans to release its Q1 2024 Security Intelligence Report on May 15 2024, which will include the first quarterly count of AI‑generated phishing attempts detected across its enterprise customers. Early test data from Microsoft’s red‑team shows that AI‑based anomaly scanners can cut the average time to flag a malicious email from four hours to under ninety minutes, an improvement of roughly 78 %.