OpenAI Bolsters Cyber Defenses

Good morning. OpenAI says its new GPT-6 Cyber model addresses the increasingly urgent need to find and fix vulnerabilities while staying within the boundaries that researchers set.

“It is clear that we have a limited window to use frontier models to secure software⁠,” the company said Thursday. The company said its goal was to expand Daybreak, its cybersecurity program that combines AI, security solutions, partners, and support.

OpenAI made an effort to address rising concerns about models that can go beyond boundaries that people set for them. Those fears have contributed to broader and more general anxieties about AI, which were a major factor in the midterm elections in which Democrats flipped control of the House and Senate.

Last month, JPMorgan Chase Chairman and CEO Jamie Dimon told WSJ Leadership Institute President Alan Murray that business leaders needed to boost their defenses against cybersecurity threats that have mounted with the arrival of powerful new AI models. “I wrote about cyber this year. I said, and I meant it, it’s our biggest risk,” Dimon said. You can watch their full discussion on the WSJ Leadership Institute Presents: Leaders podcast. Or you can listen on Apple or Spotify.

“One risk for cybersecurity agents is that they go outside the scope of their assigned task, attempt to exploit unrelated services, or disregard the researcher’s rules of engagement,” OpenAI said. “We made alignment a central focus of GPT‑6 Cyber, using specialized training to help the model stay within authorized scope, follow restrictions on its actions, and report its work accurately.”

The company said it also addressed the growing threat of people who use frontier models to cause harm. “These attempts include efforts by malicious cyber actors and adversarial groups seeking to distill our models. We have thus paired GPT‑6 Cyber’s stronger capabilities with stronger safeguards to detect and disrupt misuse and keep the model’s actions within the user’s intent and authorization,” the company said.

GPT-6 Cyber was also designed to improve the protection of critical infrastructure such as power plants and water treatment facilities and their operational technology and industrial control systems.

On ExploitGym, a benchmark that tests whether models can exploit a specified vulnerability, GPT‑6 Cyber showed a 51% relative improvement over GPT‑5.6 Cyber, according to OpenAI.

The retreat from the AI trade appears to be easing after a post-midterm selloff that put the S&P 500 on track for its worst week since last year’s “Liberation Day,” when President Trump raised the average effective tariff rate, The Wall Street Journal reports this morning. Nasdaq-100 futures are up more than 1%.

Some of Thursday’s biggest losers were bouncing back in premarket trading, including AMD, which fell 17% Thursday, along with Nvidia, CoreWeave and Broadcom, which each dropped 9% to 10%.

And then there is this. The U.S. added 165,000 jobs in October, the Labor Department said Friday, far exceeding expectations. The unemployment rate slipped to 4.1%. Economists had expected 4.2%.Now about those losses. Almost nothing in tech escaped Thursday’s market rout. Chip makers in Asia, power producers in Texas, software makers in Silicon Valley, contractors and data center developers all sank together. The Magnificent Seven lost about $1 trillion in market value, chip stocks had their worst day since the pandemic, and the Nasdaq composite entered a correction after shedding an estimated $2.5 trillion.

The selloff also raised doubts about Anthropic’s IPO. The AI lab has been aiming to go public the week before Thanksgiving, raising up to $100 billion at a valuation of about $1.5 trillion. But three major investors who plan to take part now say the valuation will probably have to come down, The Wall Street Journal reports.

Some investors want to see Anthropic’s October numbers before assuming a cut. Its annualized revenue run rate topped $80 billion at the end of the month. But some interest was cooling even before Thursday: rival OpenAI has raised less than half of the $30 billion it targeted in its latest round, at a lower valuation than it had sought.

With the company’s business deeply tied to the biggest tech players, the stakes reach beyond Anthropic, the WSJ notes. Investors in both public and private markets fear that a change in Anthropic’s plans could hit the rest of their holdings.

Candidates who promised to freeze data center development won gubernatorial races in Texas, Georgia, Ohio, and Michigan, all major data center hubs. Voters also approved a host of ballot measures aimed at curbing construction, leaving AI companies scrambling.

At OpenAI, Sam Altman fielded employee questions about the backlash, and senior leaders debated whether to lower next year’s capacity targets, The Wall Street Journal reports.

At OpenAI, Sam Altman fielded employee questions about the backlash, and senior leaders debated whether to lower next year’s capacity targets, The Wall Street Journal reports. Across the industry, government relations teams rushed to connect with incoming governors.

Some companies are looking overseas. Meta Platforms President Dina Powell McCormick said at this week’s WSJ Tech California event that Meta and other tech companies are increasingly considering building new data centers outside the U.S. to get around the new hurdles.

Others are seeking more wiggle room at home:

The WSJ Tech Council brings together CIOs, CTOs and CISOs advancing innovation and shaping the future. Join this trusted community where tech executives connect with peers to explore emerging trends and gain the perspective they need to stay ahead of disruption.

Request Information

Follow Isabelle Bousquette on LinkedIn, Instagram, X, and TikTok for more behind the scenes on her tech and AI coverage.

Follow Belle Lin on LinkedIn and X for her latest reporting on enterprise technology and AI.

Steven Rosenbush is chief of the enterprise technology bureau at the WSJ Leadership Institute. He also has a column. You can follow him on LinkedIn.

Tom Loftus is the editor of The Morning Download. He suggests following Isabelle, Belle and Steve on their various social channels. But if you insist, here’s his LinkedIn.