Artificial intelligence news.
Straight from the original announcements.

Berkeley CLTC Report Warns AI Is Compressing Cybercrime Response Times

November 19, 2026, 18:15By the AI News Bites Editorial TeamSource: Berkeley CLTC
Berkeley CLTC Report Warns AI Is Compressing Cybercrime Response Times
Image: Berkeley CLTC

A new Berkeley Center for Long-Term Cybersecurity white paper says artificial intelligence is making familiar cybercrime techniques faster, cheaper and more convincing. The report urges organizations, governments and regulators to strengthen verification, governance and coordination as attacks increasingly exploit trust and time pressure.

Key points
  • A CLTC white paper examines how AI is changing cybercrime techniques.
  • A UC Berkeley exercise involved 22 participants designing and responding to AI-enabled attacks.
  • Voice cloning, deepfakes and tailored phishing lures can be produced with little skill or money.
  • The report recommends stronger verification, training, monitoring and information sharing.
  • Governments are urged to update legal frameworks and expand cooperation.

Tabletop exercise examined AI-enabled attacks

When Trust Becomes an Attack Surface: What a Tabletop Exercise Reveals about AI-Enabled Cybercrime was authored by Gil Baram, a Non-Resident Research Fellow at CLTC and Visiting Associate Professor at the Bellini College for Artificial Intelligence, Cybersecurity and Computing at the University of South Florida.

The report is based primarily on a panel and scenario-based tabletop exercise held at UC Berkeley in December 2024. The exercise examined how AI changes incentives for attackers, affects organizational responses and creates governance challenges.

A group of 22 cybersecurity practitioners, law enforcement officials, industry executives, and representatives from academia and civil society took on the roles of two criminal organizations, one state-backed and one independent. They designed AI-enabled attack campaigns, then switched to defender roles and responded to the consequences, including the compromise of a managed security service provider that cascaded into clients’ networks.

The Berkeley exercise was part of workshops conducted in partnership with the Berkeley Risk and Security Lab and Fortinet. Other workshops took place in Singapore and Tel Aviv, Israel.

AI expands the reach of familiar deception

Baram’s findings describe AI as an amplifier of existing fraud and impersonation methods, including phishing and ransomware attacks. Voice cloning, deepfake video and localized phishing lures can now be produced with little skill or money, while a single actor can generate thousands of tailored lures.

The report says this volume shifts the economics of cybercrime toward attackers. Repetition can increase the chances of success, overwhelm human review and bypass conventional filtering. Leased AI toolkits also allow people already involved in other illicit activity to enter cybercrime with minimal financial cost, skill or risk.

AI creates additional risks inside organizations. Synthetic résumés and real-time deepfake interviews can help an impostor get hired, particularly when recruitment and onboarding take place remotely. Attribution also becomes harder because fluent AI-written text removes linguistic clues and synthetic media can mislead both victims and investigators.

Recommendations focus on verification and coordination

The paper identifies recurring failure modes involving identity verification, executive decision-making under time pressure and coordination across organizational boundaries. It recommends AI-based detection and continuous monitoring systems, stronger verification processes, multi-factor authentication and confirmation of sensitive requests through a separate, pre-established channel, such as a call to a known caller.

Organizations are also advised to invest in sustained workforce training, conduct joint drills and participate in information-sharing initiatives. For governments and regulators, the report calls for mandatory incident reporting and intelligence sharing, expanded law enforcement capacity, updated legal frameworks addressing deepfakes and other AI misuse, and stronger international and public-private cooperation.

“AI has made familiar forms of deception faster, cheaper, and more convincing,” Baram writes in the report. He says compressed verification timelines turn “trust, authority, and coordination into primary attack surfaces.” The findings suggest that users and organizations may need to make high-risk decisions more carefully, while responding within shorter windows as AI-enabled deception becomes more effective.

Related news

Military AI Faces New Risks as Adversaries Target Its Data
Safety and ethicsNovember 19, 2026, 21:00

Military AI Faces New Risks as Adversaries Target Its Data

Adversaries could target and manipulate the data used by artificial intelligence in military decision-making, creating risks of deadly miscalculations, according to…

AI Cybercrime Turns Trust and Verification Into Primary Attack Surfaces
BusinessNovember 19, 2026, 17:16

AI Cybercrime Turns Trust and Verification Into Primary Attack Surfaces

A new Berkeley CLTC white paper finds that artificial intelligence is making phishing, ransomware, fraud, and impersonation faster and more convincing. Its central…

Julius Graf named Berkeley’s 2026-27 Cal Cybersecurity Research Fellow
Safety and ethicsNovember 19, 2026, 00:45

Julius Graf named Berkeley’s 2026-27 Cal Cybersecurity Research Fellow

UC Berkeley has named Julius Graf its 2026-27 Cal Cybersecurity Research Fellow, supporting research into security threats involving generative AI and autonomous agent…