LATEST: OpenAI says its upcoming model Astra may have reached "critical" cybersecurity capabilities, including the ability to build zero-day exploits without human intervention.

What happened

OpenAI announced that its upcoming AI model, named Astra, may have achieved "critical" cybersecurity capabilities, specifically the ability to automatically create zero‑day exploits—software vulnerabilities that are unknown to defenders—without any human involvement.

Confirmed

Global impact / market context

If true, Astra could lower the barrier for attackers to develop powerful, undetectable malware, increasing cyber risk for businesses and governments. It also raises ethical questions about releasing advanced AI tools that can be weaponised.

Analyst inference

The claim comes as AI firms race to showcase more capable models, while regulators and security firms warn about AI‑driven threats. Investors are watching how such announcements affect valuations of AI and cybersecurity companies.

Analyst inference

What to watch

1. OpenAI’s detailed technical disclosures on Astra’s capabilities, which would confirm or clarify the extent of its zero‑day generation ability. Proposed
2. Responses from cybersecurity firms and government agencies, indicating whether they see an immediate need to adjust defenses or policy. Analyst inference
3. Any regulatory actions or industry guidelines targeting AI models that can create exploits, which could shape future development and deployment rules. Proposed

Evidence

• x.com