Wesley Bull

Assessing the security risks of the world’s most iconic people and companies, revealing the gaps they don’t know exist. | Ex-NVIDIA | Spear’s 500 Adviser

[FBI Director Kash Patel @FBIDirectorKash]

Guilty - California mayor Eileen Wang just pled guilty this week after an @FBI investigation revealed she was covertly using her position to act as an agent of China. She’ll now face justice.

This FBI and @TheJusticeDept continue moving aggressively to root out foreign agents operating in America. https://t.co/QTSfZizjVf

[J Michael Waller @JMichaelWaller]

Former head of Harvard chemistry department defects to China, tied to helping CCP build AI supersoldiers.

[Ann Srivastava @helloparalegal]

Most lawyers I talk to are using https://t.co/ytVmjMaok5 for client work.

That's a problem most don't know they have.

There's a second product - Claude Code that runs on the Anthropic API directly. The confidentiality difference between the two for legal work is not marginal. It's structural.

And in September 2025, the gap got a lot wider.

Three things to know:

→ https://t.co/ytVmjMaok5 (Free, Pro, even Team): As of late 2025, Anthropic uses your conversations to train models unless you actively opted out when prompted. Retention for opted-in users extends to 5 years.

→ Claude Code (and any API-direct integration): No training. Ever. Inputs and outputs are deleted within 7 days. The Commercial Terms of Service contractually prohibit Anthropic from training on your data.

→ An option that does not exist on https://t.co/ytVmjMaok5: Enterprise API customers can sign a Zero Data Retention agreement. Under ZDR, no data is stored at rest after the API response is returned, except for safety violations. There is no ZDR equivalent on https://t.co/ytVmjMaok5.

The question every lawyer running client work through Claude right now should be asking:

"Did I opt out of training on my https://t.co/ytVmjMaok5 account when prompted, or did I tap through it?"

If you don't know - your privileged client communications may be sitting in a 5-year training pipeline.

You probably can't run Claude Code yourself. That is fine.

The point is: when a legal tech vendor pitches you "Claude-powered AI," ask one question.

"Are you running on the Claude API directly, or through the https://t.co/ytVmjMaok5 web interface?"

The answer tells you what infrastructure your privileged client data is running on.

If you don't know - that's the diligence call you make this week.

[Chris Ryan @Watchdog_MP]

🚨 BREAKING: Toronto Police just seized “SMS Blasters” fake cell towers never seen before in Canada.

These portable devices hijack thousands of phones at once, blast fake bank/Canada Post texts, and knock out real service (even 911 calls).

Tens of thousands of phones hit.

Over 13 MILLION disruptions.

Three men charged 🇨🇳
• Dafeng Lin, 27, of Hamilton
• Junmin Shi, 25, of Markham
• Weitong Hu, 21, of Markham

This is next-level cyber crime on our streets. Stay alert. Never click surprise links.
#Toronto #CyberCrime #ScamAlert

[Wesley Bull @OnSecurityRisks]

@Jason I’m in. And all-in.

[Wesley Bull @OnSecurityRisks]

Excellent analysis 👇🏻

[Gerald Posner @geraldposner]

AI doesn’t create shortcuts to excellence. It exposes the absence of rigor. The question isn’t whether AI is powerful. It’s whether the professionals using it still are.

When I left law school, I worked at one of New York’s legendary firms, Cravath, Swaine & Moore. One of our fiercest rivals was Sullivan & Cromwell. Sloppiness wasn’t just frowned upon—it was a firing offense.

Fast forward to 2026: clients are paying up to $3,000 an hour for lawyers to file briefs filled with fake case citations generated by AI “hallucinations.” The errors weren’t caught internally. They were identified by opposing counsel. S&C had to apologize to a federal judge.

In my day, an associate who submitted invented citations would be gone by lunchtime. No debate.

This isn’t just about one firm. It marks a transition point. What once required painstaking verification—checking every citation, confirming every source—can now be generated instantly. And if that output isn’t rigorously reviewed, errors don’t just slip through. They multiply, often with a false veneer of authority.

In earlier eras, the failure point was human fatigue or oversight. Today, it’s overreliance on systems that create the illusion of competence.

AI doesn’t make professionals better. It makes their habits more visible.

Paywall-free 🔗
https://t.co/L5mcfCEt5K

[Financial Times @FT]

The firm, whose partners bill more than $2,000 per hour in bankruptcy cases, apologised for multiple AI-generated 'hallucinations' in a high-profile case. https://t.co/H0jDKbKX8E

[Scott Stewart @stick631]

Another incident at Sam Altman's residence. The celebratory and supportive commentary on social media surrounding these attacks has been horrific.

Other people in the AI space should take note of this sentiment and increase their situational awareness and security posture.

[Financial Times @FT]

Sam Altman’s San Francisco home was the target of a Molotov cocktail attack on Friday, with the suspect arrested outside the company’s office after threatening to burn it down. https://t.co/innfSowJUj

[Wesley Bull @OnSecurityRisks]

This isn’t just legal — it’s the whole family office model shifting.

The smartest families are moving to hub-and-spoke: small core team, fractional senior talent on demand. Security, risk, intelligence — same trend.

Ackman isn’t setting a trend. He’s confirming one.

[Mr Family Office @MrFamilyOffice]

Bill Ackman replaced his $1M family office legal counsel with a fractional alternative

This signals a wider trend in the family office world

Families are embracing a hybrid / virtual models

Fractional talent
Small central teams
Lean costs

[zerohedge @zerohedge]

*JONES DAY LAW FIRM SAYS IT WAS BREACHED, HACKERS STOLE DATA

*JONES DAY SAYS ATTACKERS ACCESSED FILES FROM 10 CLIENTS

[POLITICOEurope @POLITICOEurope]

EXCLUSIVE: We've obtained a cache of text messages and recorded conversations between Russian intelligence officers and a student they coerced into becoming an informant.

Here’s a rare insight into Moscow’s efforts to infiltrate opposition groups abroad: https://t.co/dhUE8JDSH3

[Intel Takes @inteltakes]

Every chief of station reading this.

[Valerie Anne Smith @ValerieAnne1970]

"Chinese 'kill switches' found hidden in U.S. solar farms."
~Michael Shellenberger

"Chinese government cellular radios activated remotely can cripple power grids & threaten national security."

[Wesley Bull @OnSecurityRisks]

Proud to have contributed to the executive protection community and this worthy cause supporting @StJude #bodyguardsforkids

[TravelGov @TravelGov]

Hong Kong: On March 23, 2026, the Hong Kong government changed the implementing rules relating to the National Security Law. It is now a criminal offense to refuse to give the Hong Kong police the passwords or decryption assistance to access all personal electronic devices including cellphones and laptops. This legal change applies to everyone, including U.S. citizens, in Hong Kong, arriving or just transiting Hong Kong International Airport. In addition, the Hong Kong government also has more authority to take and keep any personal devices, as evidence, that they claim are linked to national security offenses. Read more: https://t.co/K5w2tETFu5

[Los Angeles Times @latimes]

What began as a routine check triggered by a persistent odor led to an unsettling discovery: a hidden lab operating inside a California warehouse containing dangerous pathogens including HIV, malaria, COVID-19 and Ebola.

https://t.co/zOW3SSPO6U

[Rob Freund @RobertFreundLaw]

Message from the CA State Bar just now.

“Attorneys must independently verify any AI-assisted work product before relying on it in any context.”

[INTERPOL @INTERPOL_HQ]

Fraud is no longer a peripheral threat, it is at the centre of polycriminality, intersecting with organized crime, human trafficking and cybercrime.

🔎 The 2026 INTERPOL Global Financial Fraud Threat Assessment, released today at the #GlobalFraudSummit, highlights:

🔵 The increased use of AI to plan and execute fraud campaigns
🔵 The systematic integration of sextortion into scams
🔵 The use of specialized money laundering groups
🔵 The globalization of trafficking-fuelled scam centres

The good news? Law enforcement authorities are collaborating more effectively.

Read more: https://t.co/HGSDhnUVC0

[Corporate Counsel @CorpCounsel]

The Rise of Deepfakes: What to Know and How to Prepare https://t.co/cR0a5urFkL