OpenAI Doubles Down on Cybersecurity With $1B Investment and Zero-Day AI

OpenAI is putting $1 billion toward subsidized access to its cybersecurity models over the next six months, following the disclosure that its Astra system can discover previously unknown software flaws and convert them into working exploits without requiring continuous human direction.

The company said Thursday that the program is designed to help organizations defend against increasingly sophisticated AI-driven cyberattacks, including zero-day exploits that could be produced by future AI models.

Zero-day attacks exploit software weaknesses that developers have not yet identified or patched. Because no fix is available when the vulnerability is exploited, such attacks can be particularly difficult to stop.

OpenAI said the $1 billion commitment is not a grant or dedicated investment fund. Instead, organizations will receive discounted access to Daybreak, the cybersecurity platform introduced by the company earlier this year, along with technical assistance and training.

Daybreak is available in two tiers. The Blue tier uses OpenAI’s standard models for everyday defensive cybersecurity tasks, while the Red tier provides approved organizations with specialized models for more sensitive security operations. OpenAI said roughly 2,000 organizations currently use the platform.

The initiative will prioritize critical infrastructure and public-interest groups, including water and wastewater operators, electricity-grid companies, state and local governments, community and regional banks, nonprofits and open-source software maintainers.

Although crypto exchanges, custodians and blockchain protocols were not specifically mentioned, open-source projects supporting the crypto industry could fit within the broader target group.

Bitcoin Core, Ethereum clients and libraries used by DeFi protocols are often maintained by small teams and volunteer developers, despite the software helping safeguard billions of dollars in digital assets.

Astra Changes the Cybersecurity Equation

OpenAI’s announcement follows its disclosure earlier this week that Astra can independently locate previously unknown vulnerabilities and develop functional attacks from them.

The capability could mark a significant shift in cybersecurity. Finding and exploiting complex software weaknesses has traditionally required experienced researchers and specialized hacking groups. AI systems capable of automating much of that process could lower the barrier to sophisticated attacks.

Organizations that qualify can apply for subsidized Daybreak access through the platform’s website. OpenAI also plans to extend the program to partner countries in the coming weeks.

Crypto Software Faces New Vulnerability Risks

The announcement comes after several security incidents exposed weaknesses in widely used crypto infrastructure.

In August, a flaw in BTCPay Server, an open-source application used by merchants accepting Bitcoin payments, exposed credentials controlling Lightning nodes. Attackers exploited the vulnerability and drained funds before a patch was released.

Core Lightning, a major implementation for operating the Bitcoin Lightning Network, subsequently advised users to disconnect their machines after AI-generated vulnerability reports identified several real security issues.

Coldcard, a hardware wallet for offline Bitcoin storage, also released new firmware following a $114 million theft. The company said frontier AI models helped uncover separate bugs that were not related to the original attack.

The incidents have increased concern within the crypto industry that AI could strengthen attackers faster than it improves the defensive capabilities available to developers.

More than 30 companies, including Coinbase, Block, BitGo, Blockstream and ARK Invest, have signed an open letter asking AI laboratories to give security teams early access to their most powerful models.

The companies said AI labs were screening organizations before granting access to their strongest systems, while attackers could potentially gain access to similar capabilities through other channels.

That could leave developers maintaining Bitcoin and other critical open-source infrastructure at a disadvantage, forcing them to protect billions of dollars in assets with less capable AI tools than those potentially available to attackers.

OpenAI’s Daybreak initiative is intended to narrow that gap by expanding access to advanced AI systems for defensive cybersecurity work as automated vulnerability discovery becomes more capable.