The Justice Department has launched a probe into Andreessen Horowitz regarding the firm’s partners serving on the boards of competing companies, Bloomberg reported.

The nearly year-long investigation focuses specifically on the firm’s board seats at Databricks, which is valued at $190 billion, and Fivetran, which combined with dbt Labs in June. The firm’s co-founder, Ben Horowitz, serves on the board of Databricks, while partner Martin Casado serves on the board of Fivetran.

Several VCs told TechCrunch they were surprised by news of the probe. Databricks and Fivetran are competitors now, but the two companies weren’t rivals when a16z invested in the startups, according to another Databricks investor who spoke on condition of anonymity. Databricks is largely known for its cloud storage products but, with its Lakeflow product, has expanded into AI data pipelines and application connectors. That’s Fivetran’s main business.

Given that Andreessen Horowitz has backed hundreds of companies, it’s almost inevitable that some startups will pivot or expand into the same markets, becoming competitors.

While backing direct rivals has become more acceptable recently, as evidenced by the many VCs that funded both Anthropic and OpenAI, holding a board seat on competing startups creates a far greater conflict of interest. Directors are generally privy to much more sensitive strategic information than non-board investors ever see.  

Such conflicts can be resolved by having a partner step down from one of the boards. However, because Databricks and Fivetran have different individuals from the same VC firm on their boards, a16z can institute a so-called Chinese wall between Horowitz and Casado, which would prevent the two partners from sharing confidential information about the two companies with each other, one investor said.

The investigation invokes Section 8 of the Clayton Act, a 112-year-old law stating that an individual or entity is barred from serving on the boards of competing companies. Since regulators have rarely targeted venture capital with this rule, the industry is watching the DOJ’s probe closely. If a16z is forced to surrender a seat, founders may place less value on board commitments from top-tier VCs, given that those investors might be forced to step down if a portfolio overlap creates a future conflict.

a16z did not immediately respond to our request for comment, nor did it respond to Bloomberg. Databricks and DOJ declined comment.

TikTok is developing a feature that would allow users to send each other money via direct messages, according to a new report from Bloomberg. If rolled out, the feature would use the social media service’s TikTok Pay offering, which is already available in Southeast Asia for TikTok Shop purchases.

References to the potential feature were found in code hidden within the current version of TikTok’s U.S. iPhone app, according to the report. The code indicates that recipients would be able to “tap to accept” payments, while senders could include messages with their payments, similar to Venmo.

TikTok told Bloomberg that the feature is not being tested, which suggests that it’s in early development. Given that the feature is still under development, it’s unknown when or if TikTok plans to widely release peer-to-peer payments.

TikTok did not immediately respond to TechCrunch’s request for comment.

It’s worth noting that this isn’t the first time TikTok has tried to push further into financial services. Reuters reported earlier this year that TikTok had applied to Brazil’s central bank for approval to operate as a financial technology company offering lending and payment services. 

Although TikTok is widely described as a social media giant, it has gradually expanded beyond that category thanks to additions such as robust search, TikTok Shop, a local discovery map, games, hotel bookings, and more. By introducing peer-to-peer payments, TikTok would be competing with services like Venmo and Zelle.

TikTok isn’t the only social network pushing into financial services, as X, formerly known as Twitter, recently launched X Money to allow users to send each other money.

On Tuesday, OpenAI announced a new batch of new security policies focused on containing security incidents while models are being tested. The new safeguards include more detailed monitoring of models during the development process, as well as greater emphasis on alignment and security during the post-training process.

“As models become more capable, the risks associated with developing and testing them internally also grow,” the company said in a blog post. “Our standards for monitoring, alignment, and security must stay ahead of those risks.”

The new measures are one of the first public changes in OpenAI’s safety practices since the immediate aftermath of the Hugging Face incident, which was disclosed on July 26th.

OpenAI representatives emphasized that the measures are not a direct response to the Hugging Face incident, but were also provoked in part by the cybersecurity capabilities of the forthcoming Astra model, as well as the overall pace of progress in AI development.

In the same post, OpenAI disclosed that it had freezed reinforcement learning for two weeks following the Hugging Face incident, but had since restarted many of the less risky models.

“Our largest planned frontier RL run remains on hold while we conduct smaller-scale training and evaluations to assess model behavior, validate our safeguards, and establish more evidence of alignment before proceeding,” the post reads.

Speaking to reporters, OpenAI’s VP of research Amelia Glaese emphasized that the strictness of the controls would increase as models became more capable, with the largest models facing the greatest scrutiny.

“We have put in place requirements and expectations for safe development,” Glaese told reporters. “Those requirements and expectations vary with the level of risk that we that we see.”

OpenAI has been criticized for poor network security practices in the wake of the incident, which saw models escape their training environment by compromising a packet-installation utility that retained access to the internet. The new safeguards include stronger network isolation practices, although the specifics remain vague. Under the new system, the post says, “a single compromise of a workload or supporting service does not, by itself, allow for unauthorized access to the Internet, or other internal networks.”

The strongest safeguard is the monitoring system, which will examine tool actions, available reasoning traces and activity logs for a variety of unauthorized behavior. OpenAI says they aim to issue alerts within 30 minutes of the concerning activity.

OpenAI estimates that the compute burden of that monitoring will be roughly 20% of whatever process is being monitored. The company promised further details on the system in a forthcoming blog post. OpenAI’s official post-mortem analysis of the event is also still pending.