OpenAI, following Anthropic, has released an AI for finding holes in software. Should we expect "bugmageddon"?

OpenAI has opened up access to a new neural network capable of identifying vulnerabilities in enterprise software to a narrow range of customers. A week earlier, a similar release by competitor Anthropic crashed the shares of cyberthreat defense software developers, prompting Wall Street to talk about a new threat to tech sector quotes - "bugmageddon.

OpenAI's neural network, called GPT-5.4-Cyber, is designed to find loopholes in program code. It alerts information security professionals to fix problems before they are exploited by attackers. OpenAI announced the product launch on April 14, a week after Anthropic gave a small number of organizations access to its Mythos AI model with similar functionality. Both Anthropic and OpenAI call their developments a "new line of defense" against hackers, who themselves search for vulnerabilities using neural networks, the Financial Times writes.

To extend the capabilities of GPT-5.4-Cyber, the neural network was trained with fewer constraints than OpenAI's publicly available models. OpenAI's previous product in this area, Codex Security, released in March, has already helped fix more than 3,000 vulnerabilities, the company said. To prevent GPT-5.4-Cyber from being used for illegal purposes, the neural network is still only available to members of the "top tier of the trusted access program," but that will expand "as much as possible" in the future, OpenAI promised.

Late last week, quotes of leading companies in the information security sector fell due to traders' fears that Anthropic's Mythos neural network could reshape the industry. On Friday, April 10, a basket of shares of U.S. software developers from Goldman Sachs lost 5%, the FT reported. Shares of cybersecurity specialist CrowdStrike lost 4%, Palo Alto Networks fell 7% and shares of cloud monitoring platform developer Datadog fell 3%. Quotes of funds investing in the software sector, also sagged: the securities of Ares Management fell by 4%, and Blackstone - by 2%.

Unlike the Anthropic product, yesterday's release of the GPT-5.4-Cyber model was taken calmly by investors. In the first minutes of preliminary trading on Wall Street on April 15, quotations of profile issuers showed about zero dynamics. CrowdStrike added 0.2%, Datadog - 0.4%, and Palo Alto Networks - decreased by 0.5%. The price of shares of Ares Management did not change, Blackstone - grew by 0.4%.

Three days before the Mythos neural network crashed the market, Goldman Sachs Research's chief global equity strategist Peter Oppenheimer published a new review. In it, the analyst said the tech stocks' early 2026 drawdown below the rest of the market created "attractive opportunities for investors."

Goldman Sachs bases its investment case on the fact that the IT sector's fundamentals remain strong: return on equity remains high and earnings forecasts are being revised upward more aggressively than in other industries. The bank expects returns from AI investments to drive about 40% of earnings per share (EPS) growth in the S&P 500 index this year. Barring global liquidity shocks, current low multiples on the back of high growth rates make technology stocks one of the best investments in the market, the bank pointed out.

This article was AI-translated and verified by a human editor