The rise of artificial intelligence is reshaping how crypto security works, but not always for the better. While AI is helping uncover vulnerabilities faster, it is also flooding teams with low-quality reports that are becoming harder to manage.

Bug Bounty Submissions Surge Across Crypto

Bug bounty programs, which reward ethical hackers for identifying vulnerabilities, have long been a key part of crypto security. Now, AI tools are accelerating that process by scanning large amounts of code in seconds.

According to HackerOne, there were 85,000 valid bug bounty submissions in 2025, marking a 7% increase from the previous year.

However, the total number of submissions, including invalid ones, has grown much faster.

AI Creates More Noise Alongside Value

Industry leaders say AI is driving both progress and problems.

Barry Plunkett, co-CEO of Cosmos Labs, revealed that their program has seen a 900% increase in submissions, with teams now handling between 20 and 50 reports per day.

While some of these reports are legitimate, many are not.

Kadan Stadelmann, CTO of Komodo Platform, noted a clear rise in:

• False positives
• Low-quality submissions
• Reports likely generated using AI tools

The underlying issue is simple: AI has significantly reduced the effort required to generate a bug report, leading to a flood of submissions.

Developers Struggle With “AI Slop”

The growing volume of poor-quality reports is becoming a major burden.

Daniel Stenberg, creator of the widely used curl tool, recently shut down his bug bounty program altogether, citing exhaustion from dealing with what he described as “AI slop.”

For many teams, separating real vulnerabilities from noise is now one of the biggest challenges in maintaining security.

Adapting to a New Reality

To manage the surge, crypto teams are starting to rethink how bug bounty programs operate.

Some of the strategies being implemented include:

• Prioritizing submissions from trusted researchers
• Tightening scoring and validation criteria
• Using advanced triage systems

These changes aim to ensure that critical vulnerabilities are not overlooked amid the growing volume of reports.

AI Could Also Be the Solution

Despite the challenges, AI may also help solve the very problem it created.

Developers are exploring defensive AI systems that can:

• Filter incoming reports
• Identify high-quality submissions
• Reduce the burden on human reviewers

This approach could be especially important for smaller teams with limited resources.

A Turning Point for Crypto Security

Bug bounty programs remain essential for securing decentralized systems, but the rise of AI is forcing a shift in how they are managed.

The industry now faces a balancing act:

• Leveraging AI to improve security
• Preventing it from overwhelming systems with low-quality data

As AI continues to evolve, so too will the tools and strategies needed to keep crypto protocols secure.

Privacy-focused crypto protocol Umbra has temporarily taken its front-end interface offline in an effort to slow down hackers attempting to move stolen funds.

The move comes amid heightened scrutiny following a series of major exploits across the crypto ecosystem.

Front-End Taken Offline After Suspicious Activity

Umbra said it identified roughly $800,000 in stolen funds being routed through its protocol. In response, the team placed its hosted front end into maintenance mode.

The protocol noted that the interface will remain offline until it is confident that restoring it will not interfere with ongoing recovery efforts.

This action follows the recent exploit of Kelp DAO, where attackers stole over $280 million, with some reports linking the movement of funds through Umbra.

Limits of Control in Decentralized Systems

Despite shutting down its front end, Umbra acknowledged a key limitation: it cannot stop users from interacting directly with its smart contracts.

Because the protocol is open-source:

• Users can access it through self-hosted interfaces
• Alternative front ends can be deployed independently
• Smart contracts remain fully operational onchain

This highlights the broader challenge of controlling decentralized infrastructure once it is live.

Debate Over Responsibility Intensifies

The situation has reignited debate around developer responsibility in decentralized systems.

Roman Storm, co-founder of Tornado Cash, argued that disabling a front end may not be enough to satisfy regulators.

Storm, who was previously convicted in a high-profile case, said authorities may still view control over a user interface as control over the protocol itself.

He warned that:

• Modifying or shutting down a front end could be interpreted as governance authority
• Developers may still face legal accountability regardless of decentralization claims

Umbra Defends Its Design

Umbra pushed back on claims that its protocol is useful for laundering funds.

The team emphasized that:

• The protocol primarily protects the receiver’s identity, not the sender’s
• Transactions remain traceable onchain
• Stolen funds routed through Umbra can still be identified

It also confirmed that it is working with security researchers to track suspicious activity.

Ongoing Pressure on Privacy Tools

The incident reflects growing pressure on privacy-focused crypto tools as regulators and law enforcement target illicit fund flows.

While some platforms have taken steps to freeze or block hacker activity, decentralized protocols like Umbra face structural limitations in enforcement.

A Balancing Act Between Privacy and Security

Umbra’s decision underscores a broader tension in crypto:

• Preserving user privacy
• Preventing misuse by bad actors

As exploits continue and scrutiny increases, protocols may face tougher choices around how much control they can or should exert over their systems.

A major Wall Street law firm has admitted fault after submitting a legal filing filled with errors caused by artificial intelligence, raising fresh concerns about AI use in high-stakes professional work.

AI Hallucinations Lead to Dozens of Errors

Sullivan & Cromwell issued a formal apology to a federal judge after a court document included around 40 incorrect citations and related mistakes.

Andrew Dietderich, co-head of the firm’s global restructuring team, acknowledged the issue in a letter to Martin Glenn of the US Bankruptcy Court for the Southern District of New York.

“We deeply regret that this has occurred,” Dietderich said, taking responsibility for the filing and confirming that the firm failed to meet required accuracy standards.

Internal Safeguards Were Not Followed

Dietderich explained that the firm already has internal policies governing AI use, including:

• Verifying citations generated by AI tools
• Reviewing all legal submissions before filing

However, in this case, those safeguards were not properly followed. As a result, the review process failed to catch both AI-generated inaccuracies and additional manual errors.

Growing Problem Across Legal Industry

The incident highlights a broader issue with AI “hallucinations,” where tools generate false or misleading information that appears credible.

According to legal technologist Damien Charlotin, there have been:

• Over 1,300 recorded AI-related errors in legal filings globally
• More than 900 cases in the United States alone

Most of these involve fabricated or incorrect legal citations, though flawed legal arguments have also surfaced.

Firm Launches Internal Investigation

Sullivan & Cromwell has since taken immediate steps to address the issue, including:

• Conducting a full internal review
• Assessing whether stronger safeguards are needed
• Considering updates to training and oversight processes

The firm also reached out to Boies Schiller Flexner LLP, which first flagged the errors, to acknowledge and apologize for the mistake.

AI Use in Law Under Scrutiny

The situation adds to growing scrutiny over how AI is being used in professional environments, especially in fields like law where accuracy is critical.

While AI tools can improve efficiency, this case shows that without strict oversight, they can introduce serious risks.

A Reminder of Human Accountability

Despite increasing reliance on AI, the responsibility for accuracy still lies with professionals.

As Dietderich noted, ensuring the reliability of legal filings is ultimately a human obligation, regardless of the tools used.