OpenAI details new actions to disrupt malicious uses of AI

OpenAI has published an update on how it is combating the malicious use of AI, outlining recent disruptions of networks involved in scams, cyberattacks and covert influence operations. The company says it has disrupted 40+ coordinated networks since early 2024 and continues to tighten policy enforcement, account bans and cross‑industry collaboration.

The report highlights work to identify and remove actors abusing AI models for phishing, fraud, malware assistance and influence campaigns, as well as investments in safety tooling and partnerships with platforms and security teams.

• Threat focus: scams, cyber intrusion assistance, disinformation/influence ops and authoritarian abuses.
• Actions taken: account takedowns, policy enforcement, model and API safeguards, and sharing signals with partners.
• Collaboration: coordination with researchers, platforms and civil society to spot and disrupt abuse earlier.
• Ongoing work: improving detection of misuse patterns and red‑team testing for new releases.

Why it matters: As AI capabilities expand, reducing weaponization risks—from scalable phishing to influence ops—helps preserve user trust and keeps AI accessible for legitimate research and productivity.

References:
OpenAI: Disrupting malicious uses of AI (Oct 2025)

What to watch: transparency around future takedown numbers, details on repeat‑offender networks, and whether industry data‑sharing expands to cover more threat categories.

Discussion: What safeguards (policy, product or partnership) do you think most effectively deter AI‑enabled scams and disinformation?